150 manually annotated ICA components

150 shown · 50 randomly sampled components from each model · profiles are collapsed initially

← Back to ICA Lens
GPT-2 Small · L0/C421

Key

Confidence 5/5
Component profile — C421 · layer 0 · positive tail

Logit-lens tokens · positive

#1" key"#2" keys"#3"stroke"#4" Key"#5"ring"#6"key"#7"hole"#8"Key"#9" Keys"#10"irs"

R-lens tokens · positive

#1" Key"#2" Keys"#3"key"#4" keys"#5"Key"#6"Keys"#7" KEY"#8"keys"#9" key"#10"KEY"

Skewness

+4.00

Excess kurtosis

77.50rank #320

Logcosh deviation

0.1021rank #422

Top-score occurrences · positive

  1. " key" as a user enters keystrokes, the↕ +25.75 (#1) · ⚡ 53.5%
  2. " key" you use a natural key would have the following↕ +25.62 (#1) · ⚡ 55.1%
  3. " key" is not a primary key ?↵↵A↕ +25.38 (#1) · ⚡ 52.0%
  4. " key"↵switch from SSH key-based authentication to↕ +25.00 (#1) · ⚡ 49.6%
  5. " key"gen I created a key pair.↵I↕ +25.00 (#1) · ⚡ 55.5%
  6. " key" need to specify a keyring file path,↕ +25.00 (#1) · ⚡ 56.2%
  7. " key" is not a primary key ↵↵Yes↕ +24.88 (#1) · ⚡ 53.5%
  8. " key"⌥) key, and click System↕ +24.75 (#1) · ⚡ 51.2%
  9. " key" reference grid, each key can be identified by↕ +24.62 (#1) · ⚡ 54.7%
  10. " key" sequencial unique key number for these rows↕ +24.50 (#1) · ⚡ 52.7%
  11. " key" presence of a given key (destination_↕ +24.50 (#1) · ⚡ 53.1%
  12. " key"How to simulate numeric keypad (1)↕ +24.50 (#1) · ⚡ 52.0%
  13. " key" an ID as unique key.↵In the↕ +24.50 (#1) · ⚡ 55.5%
  14. " key" that have the same key into an array,↕ +24.50 (#1) · ⚡ 53.5%
  15. " key" to bind the Enter key to this button such↕ +24.38 (#1) · ⚡ 53.9%
  16. " key" the code forms a key to the virtual world↕ +24.38 (#1) · ⚡ 52.7%
  17. " key"c) is a key component of the brain↕ +24.25 (#1) · ⚡ 49.0%
  18. " key"/set using a key to prevent that key↕ +24.25 (#1) · ⚡ 55.1%
  19. " key" a cookie with a key and value of an↕ +24.25 (#1) · ⚡ 51.2%
  20. " key" am trying to get key and value in a↕ +24.25 (#1) · ⚡ 51.2%
GPT-2 Small · L0/C459

Full

Confidence 5/5
Component profile — C459 · layer 0 · positive tail

Logit-lens tokens · positive

#1"xit"#2"full"#3"lly"#4"fill"#5"blooded"#6" complement"#7"fully"#8" fled"#9"colour"#10"ful"

R-lens tokens · positive

#1"Brave"#2" Fuller"#3"Full"#4" Merrill"#5"natureconservancy"#6"Mer"#7"fill"#8"pletion"#9" Crimson"#10" confir"

Skewness

+3.55

Excess kurtosis

70.78rank #334

Logcosh deviation

0.0937rank #460

Top-score occurrences · positive

  1. " full"<|endoftext|>Last year was full of discovery and learning↕ +27.75 (#1) · ⚡ 40.6%
  2. " full"October is finally in full swing, giving us↕ +26.12 (#1) · ⚡ 42.8%
  3. " full" days away from the full launch of Echo Combat↕ +26.00 (#1) · ⚡ 42.0%
  4. " full" particular dress features a full side cut out with↕ +25.75 (#1) · ⚡ 41.2%
  5. " full" in cyberspace full of odds and ends↕ +25.38 (#1) · ⚡ 42.4%
  6. " full" likely to launch a full-faired version↕ +25.12 (#1) · ⚡ 42.0%
  7. " full" America Society celebrates the full autumn moon each year↕ +25.00 (#1) · ⚡ 41.0%
  8. " full" so collectors must carry full protective gear as they↕ +24.75 (#1) · ⚡ 42.0%
  9. " full" has finally embraced a full rebuild, head coach↕ +24.75 (#1) · ⚡ 41.6%
  10. " full"Straight up and full disclosure. I'm↕ +24.75 (#1) · ⚡ 44.7%
  11. " full" a student teacher, full time teacher, and↕ +24.62 (#1) · ⚡ 40.2%
  12. " full" that I grew a full beard. Eight or↕ +24.62 (#1) · ⚡ 42.0%
  13. " full" have a nice day full of code :)↵↕ +24.62 (#1) · ⚡ 44.1%
  14. " full"You should use a full path, not a↕ +24.62 (#1) · ⚡ 45.5%
  15. " full" landscape. It's full of hills and↵↕ +24.62 (#1) · ⚡ 44.5%
  16. " full" Poffo came full circle backstage at September↕ +24.62 (#1) · ⚡ 44.7%
  17. " full".↵↵A full orchestra is dedicated to↕ +24.62 (#1) · ⚡ 43.8%
  18. " full". It has a full bathroom, kitchen and↕ +24.50 (#1) · ⚡ 44.7%
  19. " full" whilst still having the full experience of the greatest↕ +24.50 (#1) · ⚡ 44.1%
  20. " full"?" "Those are full of pictures of me↕ +24.50 (#1) · ⚡ 41.0%
GPT-2 Small · L0/C524

Personal

Confidence 5/5
Component profile — C524 · layer 0 · positive tail

Logit-lens tokens · positive

#1" occas"#2"iates"#3"iating"#4" prosec"#5" identifiable"#6" destro"#7"etheless"#8" eleph"#9" integrity"#10" polling"

R-lens tokens · positive

#1" Cosponsors"#2" Golf"#3"iffin"#4" Haley"#5" Griffin"#6" Nike"#7" CTR"#8" Sports"#9" Tillerson"#10" VIP"

Skewness

+1.35

Excess kurtosis

16.82rank #570

Logcosh deviation

0.0811rank #525

Top-score occurrences · positive

  1. " personal" does not expose the personal API, and so↕ +16.25 (#1) · ⚡ 17.1%
  2. " personal" the interview reflect his personal views based on his↕ +16.00 (#1) · ⚡ 16.3%
  3. " personal" to differentiate work from personal email.↵↵↕ +15.88 (#1) · ⚡ 18.1%
  4. " personal" do not contain any personal data. You can↕ +15.81 (#1) · ⚡ 17.0%
  5. " personal" terrorists would connect these personal profiles to moderation activities↕ +15.69 (#1) · ⚡ 16.7%
  6. " personal" support given is purely personal opinion.↕ +15.56 (#1) · ⚡ 16.3%
  7. " personal" be conducted by your personal gynecologist in↕ +15.56 (#1) · ⚡ 17.3%
  8. " personal" -- using their own personal strategies -- during 20↕ +15.50 (#1) · ⚡ 16.2%
  9. " personal" fascinating new world of personal responsibility for those who↕ +15.50 (#1) · ⚡ 15.6%
  10. " personal" the legislature cannot compel personal behavior.↵↵↕ +15.50 (#1) · ⚡ 16.7%
  11. " personal", tweeting, or personal writing necessarily represents the↕ +15.50 (#1) · ⚡ 16.4%
  12. " personal"↵These are my personal notes and I encourage↕ +15.50 (#1) · ⚡ 15.0%
  13. " personal" is usually a very personal item, at some↕ +15.50 (#1) · ⚡ 16.0%
  14. " personal" classification. For my personal business with no employees↕ +15.25 (#1) · ⚡ 16.5%
  15. " personal" can sometimes feel more personal since it’↕ +15.25 (#1) · ⚡ 14.1%
  16. " personal" candidates and mention their personal interests, mutual university↕ +15.25 (#1) · ⚡ 17.1%
  17. " personal" to offer a more personalised medical system.↕ +15.19 (#1) · ⚡ 16.5%
  18. " personal" best practices and share personal experiences fostering an environment↕ +15.19 (#1) · ⚡ 16.1%
  19. " personal" the cull often ensure personal details are not lodged↕ +15.19 (#1) · ⚡ 16.5%
  20. " personal" more split in their personal expectations.↵↵↕ +15.19 (#1) · ⚡ 17.2%
GPT-2 Small · L0/C642

Light

Confidence 5/5
Component profile — C642 · layer 0 · negative tail

Logit-lens tokens · negative

#1"══"#2" DRAG"#3"═"#4"lihood"#5" Lamp"#6"�"#7"��"#8"UTION"#9" Kant"#10" Straw"

R-lens tokens · negative

#1" Lamp"#2" lit"#3" cig"#4" Melanie"#5" smokers"#6" Light"#7" Lit"#8" smoker"#9" Hopkins"#10" lighter"

Skewness

-0.07

Excess kurtosis

11.17rank #622

Logcosh deviation

0.0679rank #643

Top-score occurrences · negative

  1. " light" determine whether a blue light (405 nm)↕ -15.25 (#1) · ⚡ 14.9%
  2. " light" wavelength conversion materials absorb light in a first wavelength↕ -14.69 (#1) · ⚡ 14.3%
  3. " light" been employed as a light source for a display↕ -14.50 (#2) · ⚡ 15.4%
  4. " light" microscopy images, light absorption characteristics, the↕ -14.44 (#2) · ⚡ 14.5%
  5. " light" been given the green light to feature by his↕ -14.38 (#1) · ⚡ 14.0%
  6. " light" light emitted from a light source, such as↕ -14.31 (#2) · ⚡ 15.3%
  7. " light"Install with or without light; optional light fixtures↕ -14.19 (#2) · ⚡ 12.8%
  8. " light" she strips off her light cotton lingerie in↕ -14.19 (#2) · ⚡ 14.5%
  9. " light" mask layer thereof absorbs light, larger playback energy↕ -14.12 (#2) · ⚡ 13.5%
  10. " light"herent light as a light source. Light from↕ -14.12 (#2) · ⚡ 15.4%
  11. " light"↵↵Despite a light schedule in Week 18↕ -14.06 (#2) · ⚡ 14.5%
  12. " light" line (marked in light green), which as↕ -14.06 (#2) · ⚡ 14.3%
  13. " light"layer structure, a light extraction structure for extracting↕ -14.06 (#2) · ⚡ 15.1%
  14. " light"iated with white fluorescent light. HIV infectivity↕ -14.00 (#2) · ⚡ 12.3%
  15. " light" molecules are all very light↵compared to↕ -14.00 (#2) · ⚡ 14.6%
  16. " light" is operating under a light load. As a↕ -14.00 (#1) · ⚡ 16.0%
  17. " light" a wavelength of the light source has been pursued↕ -13.94 (#2) · ⚡ 13.3%
  18. " light" disclosure relates to a light emitting device for converting↕ -13.94 (#2) · ⚡ 13.7%
  19. " light" was easily affected by light. However, the↕ -13.94 (#2) · ⚡ 13.9%
  20. " light" a steering mechanism to lighten the steering force↕ -13.88 (#2) · ⚡ 14.6%
GPT-2 Small · L1/C191

5

Confidence 5/5
Component profile — C191 · layer 1 · positive tail

Logit-lens tokens · positive

#1"″"#2"39"#3"42"#4"34"#5"′"#6"43"#7"41"#8"67"#9"69"#10"th"

R-lens tokens · positive

#1" 5"#2" Bosh"#3" neighbourhood"#4" Shin"#5" Pok"#6" Ced"#7" Raptors"#8" 6"#9" GTA"#10"5"

Skewness

+5.28

Excess kurtosis

53.48rank #398

Logcosh deviation

0.1486rank #192

Top-score occurrences · positive

  1. "5"<|endoftext|>Go 1.5 Release Notes↵↵↕ +15.50 (#1) · ⚡ 35.2%
  2. "5" is located 19.5 miles (29.↕ +15.44 (#1) · ⚡ 48.2%
  3. "5"t.co/5Ezre1↕ +15.31 (#1) · ⚡ 42.4%
  4. "5" average of 14.5 times earnings.↵↕ +15.12 (#1) · ⚡ 41.8%
  5. "5" with a 53.5 per cent advantage for↕ +15.12 (#1) · ⚡ 43.0%
  6. "5" {#s0005}↵===============↕ +15.12 (#1) · ⚡ 44.3%
  7. "5" circumference was 33.5 cm (3-↕ +15.12 (#1) · ⚡ 49.0%
  8. "5"umes evident 7.5 miles into the trop↕ +15.00 (#1) · ⚡ 42.6%
  9. "5""x1.5"↵↵Item↕ +15.00 (#1) · ⚡ 42.6%
  10. "5"A. 10:5-1 et seq↕ +15.00 (#1) · ⚡ 47.7%
  11. "5".2d 485 (2000)↵↕ +14.94 (#1) · ⚡ 36.7%
  12. "5" 2015, 7.5% of depression patients↕ +14.88 (#1) · ⚡ 45.1%
  13. "5" elevation of 51.5 m, its surface↕ +14.88 (#1) · ⚡ 42.2%
  14. "5"9, 71.5, 106.4↕ +14.81 (#1) · ⚡ 46.7%
  15. "5" affects approximately 12.5 million children and teens↕ +14.81 (#1) · ⚡ 40.8%
  16. "5"nell (16-5) limited the Red↕ +14.81 (#1) · ⚡ 38.5%
  17. "5" ) and 28.5 degrees anteversion (↕ +14.81 (#1) · ⚡ 48.4%
  18. "5"<|endoftext|>LRP5 and plasma cholesterol levels↕ +14.81 (#1) · ⚡ 35.2%
  19. "5" roughly $12.5 million earmarked to↕ +14.75 (#1) · ⚡ 40.4%
  20. "5"ed $1.5 million, showing there↕ +14.75 (#1) · ⚡ 42.8%
GPT-2 Small · L1/C284

Accented Latin characters

Confidence 4/5
Component profile — C284 · layer 1 · negative tail

Logit-lens tokens · negative

#1"�"#2"�"#3"ternity"#4"�"#5"�"#6"�"#7"�"#8"�"#9"�"#10"�"

R-lens tokens · negative

#1" Ö"#2" �"#3"�"#4" É"#5" é"#6"Í"#7" Ü"#8"á"#9"É"#10"î"

Skewness

-5.95

Excess kurtosis

93.59rank #294

Logcosh deviation

0.1276rank #285

Top-score occurrences · negative

  1. "�"Milan Dolinský↵↵Mil↕ -23.50 (#1) · ⚡ 27.9%
  2. "ä"ulkaisu kärjisti↕ -23.25 (#1) · ⚡ 31.1%
  3. "ä"<|endoftext|>Milbenkäse↵↵Mil↕ -22.75 (#1) · ⚡ 28.7%
  4. "�"as requisições de POST↕ -22.50 (#1) · ⚡ 29.9%
  5. "ä"Der Seriosität: Slot machine↕ -22.38 (#1) · ⚡ 29.5%
  6. "�"os e funções, o↕ -22.25 (#1) · ⚡ 32.0%
  7. "�"as Interacções Fundament↕ -22.12 (#1) · ⚡ 31.2%
  8. "�"oria su ciò che è↕ -22.12 (#1) · ⚡ 31.1%
  9. "�"as intenções da rec↕ -21.75 (#1) · ⚡ 32.4%
  10. "ê"<|endoftext|>Você sabe quem↕ -21.62 (#1) · ⚡ 23.1%
  11. "�" même où démarrait↕ -21.62 (#1) · ⚡ 31.1%
  12. "�"ážený pán pred↕ -21.38 (#1) · ⚡ 30.5%
  13. "ä"er married Maria Schärft in 1906↕ -21.38 (#1) · ⚡ 33.6%
  14. "�"ores concentrações em Conce↕ -21.38 (#1) · ⚡ 31.8%
  15. "�" das confusões em que↕ -21.38 (#1) · ⚡ 31.1%
  16. "�" de situações recor↕ -21.25 (#1) · ⚡ 31.2%
  17. "ä"st in der Häßlichkeit↕ -21.12 (#1) · ⚡ 34.4%
  18. "å"Fremgangsmåderne ved↕ -21.12 (#1) · ⚡ 19.5%
  19. "�"estourar balões de hid↕ -21.00 (#1) · ⚡ 30.7%
  20. "�"í za výbornou sp↕ -21.00 (#1) · ⚡ 31.4%
GPT-2 Small · L2/C46

First token after <|endoftext|>

Confidence 5/5
Component profile — C46 · layer 2 · positive tail

Logit-lens tokens · positive

#1" largeDownload"#2" Expand"#3" Finder"#4" ›"#5"zbollah"#6":"#7" GOODMAN"#8"buquerque"#9"mo"#10" Coverage"

R-lens tokens · positive

#1"\n"#2"Merit"#3"Returns"#4"Rocket"#5" Hover"#6"Batman"#7" Jets"#8"Kick"#9"​​"#10" Rebirth"

Skewness

+11.69

Excess kurtosis

199.29rank #82

Logcosh deviation

0.2074rank #47

Top-score occurrences · positive

  1. "Prem"<|endoftext|>Premier League Week 24↕ +21.88 (#1) · ⚡ 22.2%
  2. "�"<|endoftext|>// ©↕ +21.75 (#1) · ⚡ 18.6%
  3. "�"<|endoftext|>@using↕ +21.75 (#1) · ⚡ 18.6%
  4. "�"<|endoftext|><Project↕ +21.75 (#1) · ⚡ 18.6%
  5. "�"<|endoftext|>//----------------------------------------------------------------↕ +21.75 (#1) · ⚡ 18.6%
  6. "�"<|endoftext|>using System↕ +21.75 (#1) · ⚡ 18.6%
  7. "�"<|endoftext|>/*↵↕ +21.75 (#1) · ⚡ 18.6%
  8. "�"<|endoftext|># Copyright↕ +21.62 (#1) · ⚡ 18.4%
  9. "�"<|endoftext|>namespace↕ +21.62 (#1) · ⚡ 18.4%
  10. "."<|endoftext|>.↵↵.↵↕ +21.12 (#1) · ⚡ 46.5%
  11. "Super"<|endoftext|>Super secure Quantum cryptography has↕ +20.88 (#1) · ⚡ 28.1%
  12. "Super"<|endoftext|>Superoxide dismutase↕ +20.88 (#1) · ⚡ 28.1%
  13. "Traditional"<|endoftext|>Traditional knowledge learning and skill↕ +20.88 (#1) · ⚡ 26.6%
  14. "Pro"<|endoftext|>Prophylaxis of↕ +20.88 (#1) · ⚡ 28.3%
  15. "Pro"<|endoftext|>Prognostic value of↕ +20.88 (#1) · ⚡ 28.3%
  16. "Pro"<|endoftext|>Prostanoids induce e↕ +20.75 (#1) · ⚡ 28.1%
  17. "Super"<|endoftext|>Superficial Sider↕ +20.75 (#1) · ⚡ 27.9%
  18. "Pro"<|endoftext|>Procedures such as↕ +20.75 (#1) · ⚡ 28.1%
  19. "Pro"<|endoftext|>Properidine↵↵↕ +20.75 (#1) · ⚡ 28.1%
  20. "Visual"<|endoftext|>Visual stimuli that are spat↕ +20.75 (#1) · ⚡ 26.2%
GPT-2 Small · L3/C61

’

Confidence 5/5
Component profile — C61 · layer 3 · negative tail

Logit-lens tokens · negative

#1"️"#2"�"#3"Balt"#4"VICE"#5"··"#6"女"#7"½"#8"50"#9"ATH"#10"••••"

R-lens tokens · negative

#1"�"#2"�"#3"�"#4"�"#5"�"#6"�"#7"†"#8"ð"#9"�"#10"�"

Skewness

-7.94

Excess kurtosis

106.08rank #136

Logcosh deviation

0.1658rank #62

Top-score occurrences · negative

  1. "�" bright as they’ve always done.↕ -16.62 (#1) · ⚡ 56.2%
  2. "�" recent years we’ve seen moves that↕ -16.62 (#1) · ⚡ 47.7%
  3. "�" days, we’re launching an necessary↕ -16.50 (#1) · ⚡ 55.1%
  4. "�"As the game’s protagonist, you↕ -16.38 (#1) · ⚡ 54.3%
  5. "�" if you don’t want to read↕ -16.38 (#1) · ⚡ 53.9%
  6. "�"itan. He’s a neurologist↕ -16.38 (#1) · ⚡ 55.5%
  7. "�" international community wasn’t thrilled with Turkey↕ -16.38 (#1) · ⚡ 50.8%
  8. "�" the story doesn’t start there…↕ -16.38 (#1) · ⚡ 52.7%
  9. "�"↵↵We’re not here to↕ -16.38 (#1) · ⚡ 46.9%
  10. "�"<|endoftext|>You wouldn’t believe some of↕ -16.38 (#1) · ⚡ 42.4%
  11. "�" the sustainability director’s role as anticipating↕ -16.38 (#1) · ⚡ 60.5%
  12. "�"Why leaders don’t ask questions↵↕ -16.38 (#1) · ⚡ 53.9%
  13. "�" ensure sewing doesn’t become a lost↕ -16.25 (#1) · ⚡ 55.9%
  14. "�". My father’s musical tastes ran↕ -16.25 (#1) · ⚡ 59.8%
  15. "�" course we don’t do facel↕ -16.25 (#1) · ⚡ 53.9%
  16. "�". So I’m excited to pull↕ -16.25 (#1) · ⚡ 43.4%
  17. "�" age, it’s a class 5↕ -16.12 (#1) · ⚡ 47.9%
  18. "�": It didn’t seem to do↕ -16.12 (#1) · ⚡ 56.2%
  19. "�" of the writers’ studios and a working↕ -16.12 (#1) · ⚡ 64.5%
  20. "�", we don’t know, and↕ -16.12 (#1) · ⚡ 52.0%
GPT-2 Small · L3/C393

Contact relative clause / omitted “that”

Confidence 5/5
Component profile — C393 · layer 3 · positive tail

Logit-lens tokens · positive

#1" deem"#2"abouts"#3" deems"#4"ォ"#5" otherwise"#6" encount"#7" deemed"#8" crave"#9"��"#10" sorely"

R-lens tokens · positive

#1" personally"#2" ie"#3" alot"#4" explan"#5" etc"#6" whilst"#7" assum"#8" cue"#9" myself"#10" smokes"

Skewness

+3.12

Excess kurtosis

30.44rank #410

Logcosh deviation

0.0807rank #394

Top-score occurrences · positive

  1. " faces" highlighting the difficulties it faces as it moves past↕ +18.88 (#1) · ⚡ 20.2%
  2. " you"ing exactly which variety you have. Nevertheless,↕ +16.75 (#1) · ⚡ 35.9%
  3. " cite"↵The phrase you cite can be easily go↕ +16.62 (#1) · ⚡ 17.4%
  4. " she" Anthony, the woman she condescendingly referred↕ +16.50 (#1) · ⚡ 25.6%
  5. " provide" products or services you provide. Once you have↕ +16.38 (#1) · ⚡ 21.3%
  6. " gives" write the speeches he gives. They put THEIR↕ +16.38 (#1) · ⚡ 22.1%
  7. " it" of selling milk which it had purchased at wholesale↕ +16.38 (#1) · ⚡ 31.6%
  8. " wish" in any order you wish. You may choose↕ +16.38 (#1) · ⚡ 14.9%
  9. " they" Beatrice, whom they adored. My↕ +16.25 (#1) · ⚡ 30.1%
  10. " has" all the projects he has produced, and also↕ +16.12 (#1) · ⚡ 33.4%
  11. " had" use every space they had to offer to aid↕ +16.00 (#1) · ⚡ 26.2%
  12. " are" know the person you are ordering for," he↕ +16.00 (#1) · ⚡ 32.0%
  13. " they" temple, which power they can use to benefit↕ +15.94 (#1) · ⚡ 34.2%
  14. " had" plastic cup the defendant had placed on the counter↕ +15.88 (#1) · ⚡ 26.8%
  15. " are" away at who we are until more of what↕ +15.69 (#1) · ⚡ 27.7%
  16. " he" with Iran, which he warned would only pave↕ +15.62 (#1) · ⚡ 30.3%
  17. " believed" was the handgun we believed used in the robberies↕ +15.62 (#1) · ⚡ 16.1%
  18. " it" individuals and organisations that it says "spread hate↕ +15.56 (#1) · ⚡ 27.1%
  19. " are" about the medicines you are taking and any allergies↕ +15.56 (#1) · ⚡ 36.5%
  20. " could" the smallest one we could. There are other↕ +15.50 (#1) · ⚡ 23.4%
GPT-2 Small · L3/C409

Christ

Confidence 5/5
Component profile — C409 · layer 3 · negative tail

Logit-lens tokens · negative

#1" gra"#2"isphere"#3" scriptures"#4" scripture"#5"ievers"#6" worsh"#7" worship"#8" theology"#9" faith"#10"angel"

R-lens tokens · negative

#1" Christianity"#2" Bible"#3" worship"#4" church"#5" Jesus"#6"church"#7" Christians"#8" Biblical"#9"Jesus"#10" biblical"

Skewness

-2.37

Excess kurtosis

30.23rank #411

Logcosh deviation

0.0788rank #410

Top-score occurrences · negative

  1. " Christ" In light of the Christocentric fulness↕ -21.38 (#1) · ⚡ 22.8%
  2. " biblical" faith who respect the biblical teaching that marriage is↕ -20.00 (#1) · ⚡ 21.8%
  3. " Jesus" in The Church of Jesus Christ of Latter-↕ -19.75 (#1) · ⚡ 20.3%
  4. " Gospel"<|endoftext|>The Gospel of Jesus Christ -↕ -19.62 (#1) · ⚡ 21.2%
  5. " Christ" that the church that Christ established was not to↕ -19.50 (#1) · ⚡ 19.0%
  6. " Gospel" celebrity gossip than the Gospel? Why aren�↕ -19.25 (#1) · ⚡ 22.3%
  7. " God" Bible — "If God be for us,↕ -19.25 (#1) · ⚡ 19.5%
  8. " Christ" why the day of Christ’s crucifix↕ -19.25 (#1) · ⚡ 23.3%
  9. " Bible", 'including the Bible.'↵↵For↕ -19.00 (#1) · ⚡ 24.3%
  10. " Jesus" should note that the Jesus and Mo author has↕ -19.00 (#1) · ⚡ 22.0%
  11. " God" friend of mine says God will forgive me if↕ -18.88 (#1) · ⚡ 22.1%
  12. " God"It is true that God may have called you↕ -18.75 (#1) · ⚡ 20.8%
  13. " Jesus"↵At first, Jesus doesn’t↕ -18.75 (#1) · ⚡ 26.6%
  14. " Christ"'... the collective Christ-followers and↕ -18.62 (#1) · ⚡ 20.0%
  15. " Christ" you will abide in Christ and bear fruit for↕ -18.50 (#1) · ⚡ 19.1%
  16. " God" on his relationship with God. He does it↕ -18.38 (#1) · ⚡ 23.2%
  17. " Christ" Moshe believe in Christ, and Mullah↕ -18.38 (#1) · ⚡ 20.0%
  18. " God"icating the space as God's, used for↕ -18.38 (#1) · ⚡ 20.8%
  19. " God" chastisement from God—and ultimately complete↕ -18.38 (#1) · ⚡ 23.8%
  20. " Gospel"↵↵# The Gospel of Mark 1↵↕ -18.12 (#1) · ⚡ 19.0%
GPT-2 Small · L3/C455

Item

Confidence 5/5
Component profile — C455 · layer 3 · negative tail

Logit-lens tokens · negative

#1"ivity"#2"ively"#3" objects"#4"ious"#5" purchases"#6"handler"#7" items"#8"meal"#9" components"#10" handlers"

R-lens tokens · negative

#1"Item"#2" item"#3" items"#4" Items"#5" Item"#6" object"#7"Items"#8" objects"#9"item"#10"property"

Skewness

-1.74

Excess kurtosis

15.11rank #552

Logcosh deviation

0.0725rank #456

Top-score occurrences · negative

  1. " item" just like any other item that you can carry↕ -14.88 (#1) · ⚡ 22.8%
  2. " item" ship items varies by item, shipment and shipping↕ -14.81 (#1) · ⚡ 17.8%
  3. " item"If you return your item due to a fault↕ -14.75 (#1) · ⚡ 18.7%
  4. " item" just about every seasonal item.↵↵According↕ -14.62 (#1) · ⚡ 21.0%
  5. " item" whether or not an item is eligible for a↕ -14.44 (#1) · ⚡ 23.1%
  6. " items" Vegas are the important items, which are required↕ -14.25 (#1) · ⚡ 21.5%
  7. " item" to accurately represent each item. However, for↕ -14.00 (#1) · ⚡ 18.3%
  8. " Item"↵The Handy Item I Always Pack:↕ -14.00 (#1) · ⚡ 18.8%
  9. " items" – tell us which items are yours. And↕ -13.94 (#1) · ⚡ 20.6%
  10. " item" is much better with item specifications....and power↕ -13.94 (#1) · ⚡ 22.3%
  11. " item" can search either by item (shoes,↕ -13.94 (#1) · ⚡ 19.4%
  12. " items" merchandise package contains several items with Elysian branding↕ -13.81 (#1) · ⚡ 24.6%
  13. " items" conventional methods of identifying items.↵One proposed↕ -13.81 (#1) · ⚡ 25.6%
  14. " items" the surface appearance of items such as motorcycle helmets↕ -13.81 (#1) · ⚡ 28.3%
  15. " item" asked to remove the item.↵↵Virginia↕ -13.81 (#1) · ⚡ 21.9%
  16. " item" for a bedroll item that would put the↕ -13.81 (#1) · ⚡ 17.7%
  17. " item" are compared before the item analysis. Results suggest↕ -13.75 (#1) · ⚡ 17.4%
  18. " items" representative of all food items and food groups that↕ -13.75 (#1) · ⚡ 23.5%
  19. " Items" To Do Items↵? ↕ -13.56 (#1) · ⚡ 22.3%
  20. " item" each name as an item of the array until↕ -13.50 (#1) · ⚡ 21.9%
GPT-2 Small · L3/C650

Include

Confidence 5/5
Component profile — C650 · layer 3 · positive tail

Logit-lens tokens · positive

#1" prominently"#2"ttes"#3"ties"#4"clus"#5"ioned"#6"angering"#7"lees"#8" Deity"#9"oided"#10" Include"

R-lens tokens · positive

#1" inclusion"#2" Include"#3"clude"#4" prominently"#5" feature"#6"clusion"#7" Feature"#8" included"#9" include"#10"Feature"

Skewness

+0.69

Excess kurtosis

8.41rank #636

Logcosh deviation

0.0424rank #651

Top-score occurrences · positive

  1. " included" SSL documentation↵ * included with this distribution is↕ +12.94 (#1) · ⚡ 18.2%
  2. " included" license.txt file included.↵↵#↕ +12.94 (#1) · ⚡ 15.7%
  3. " included"T display systems have included various correction features to↕ +12.75 (#2) · ⚡ 18.9%
  4. " included"=25) were included in the study.↕ +12.56 (#1) · ⚡ 14.2%
  5. " included". The piece 212 included a recess 216 which↕ +12.50 (#2) · ⚡ 16.5%
  6. " feature"<|endoftext|>Not as a feature to use but to↕ +12.44 (#1) · ⚡ 12.7%
  7. " feature"One significant aspect and feature of the present invention↕ +12.38 (#1) · ⚡ 13.7%
  8. " feature". This is one feature of assisted homes that↕ +12.31 (#1) · ⚡ 12.7%
  9. " included"LGPLv3 included in the↵**↕ +12.25 (#1) · ⚡ 15.6%
  10. " featured"lectric devices have featured materials such as silicon↕ +12.25 (#1) · ⚡ 16.2%
  11. " included"based systems are also included, such as CD↕ +12.19 (#1) · ⚡ 14.8%
  12. " included" it. I personally included a bit of everything↕ +12.19 (#2) · ⚡ 15.2%
  13. " included" The 2008 survey also included more questions about sexual↕ +12.19 (#2) · ⚡ 16.2%
  14. " included"topenia were included. Peripheral↕ +12.12 (#2) · ⚡ 14.7%
  15. " included".↵↵Also included in the Game of↕ +12.12 (#2) · ⚡ 13.1%
  16. " feature"-main event will feature Rousey-killer↕ +12.12 (#1) · ⚡ 13.2%
  17. " included" (C) were included in this study.↕ +12.06 (#2) · ⚡ 14.2%
  18. " feature" on the album to feature Wynton Kelly on↕ +12.06 (#1) · ⚡ 15.4%
  19. " included"3 years) were included. In 82 %↕ +12.06 (#2) · ⚡ 13.4%
  20. " included" flaps except one included multiple perforators↕ +12.06 (#2) · ⚡ 16.3%
GPT-2 Small · L3/C728

?

Confidence 0/5
Component profile — C728 · layer 3 · positive tail

Logit-lens tokens · positive

#1"ドラ"#2"ּ"#3"覚醒"#4" Scand"#5"の�"#6"ウス"#7"stro"#8"ften"#9"の�"#10"�"

R-lens tokens · positive

#1" Bundy"#2" Edwards"#3" NASCAR"#4" Nelson"#5" Dempsey"#6" Phelps"#7" Buckley"#8" Dawkins"#9" Kirby"#10" Roberts"

Skewness

+0.68

Excess kurtosis

2.69rank #727

Logcosh deviation

0.0247rank #729

Top-score occurrences · positive

  1. " *" *↵ * @var mixed A↕ +8.75 (#2) · ⚡ 18.2%
  2. "var"↵ var v1 = reverse↕ +8.56 (#1) · ⚡ 8.3%
  3. " Rex"� for Rex Tillerson Rex Wayne TillersonGary Cohn↕ +8.50 (#4) · ⚡ 2.9%
  4. "<"">↵ <div class="-↕ +8.38 (#2) · ⚡ 10.3%
  5. " *" function.↵ *↵ * @↕ +8.25 (#2) · ⚡ 13.9%
  6. " *" results.↵ *↵ * @↕ +8.25 (#2) · ⚡ 16.3%
  7. " "↵ // EncodeRune↕ +8.12 (#1) · ⚡ 14.6%
  8. " ",↵ struct call↕ +8.12 (#1) · ⚡ 14.9%
  9. " ". Tom Martin ↵is over the↕ +8.06 (#1) · ⚡ 11.4%
  10. " ",↵ ↕ +8.00 (#1) · ⚡ 14.0%
  11. "uint" rc;↵ uint8 al[16↕ +8.00 (#2) · ⚡ 6.9%
  12. "uint"16];↵ uint8 cr, r↕ +8.00 (#2) · ⚡ 7.0%
  13. " " 2 ||↵ ovflsize↕ +8.00 (#1) · ⚡ 15.4%
  14. " *" *↵ * @var int↵↕ +7.97 (#2) · ⚡ 14.7%
  15. " *"Area'↵ * @prop {Number↕ +7.94 (#2) · ⚡ 10.5%
  16. " "help---↵ Register processes to be↕ +7.94 (#1) · ⚡ 15.0%
  17. " and" majors. But Brown and his teammates point to↕ +7.88 (#1) · ⚡ 12.1%
  18. " "\↵ domain_val(↕ +7.88 (#1) · ⚡ 16.8%
  19. " " help↵ Set logger buffer size↕ +7.88 (#1) · ⚡ 15.1%
  20. " "get{↵ CTD ret↕ +7.84 (#1) · ⚡ 15.9%
GPT-2 Small · L4/C32

Be

Confidence 5/5
Component profile — C32 · layer 4 · positive tail

Logit-lens tokens · positive

#1" able"#2" considered"#3" seen"#4" judged"#5" taken"#6" replaced"#7" treated"#8" forgiven"#9" subjected"#10"acons"

R-lens tokens · positive

#1" BE"#2" PROGRAM"#3"ecause"#4" BEFORE"#5" Beth"#6"......."#7" EntityItem"#8" able"#9"psc"#10" LIST"

Skewness

+8.19

Excess kurtosis

100.95rank #87

Logcosh deviation

0.1840rank #33

Top-score occurrences · positive

  1. " be"!↵↵To be ‘shadow kissed↕ +17.00 (#1) · ⚡ 24.3%
  2. " be".↵↵To be honest, I�↕ +16.88 (#1) · ⚡ 25.6%
  3. " be".↵↵To be honest, I've↕ +16.75 (#1) · ⚡ 25.8%
  4. " be" the fall. To be fair, some of↕ +16.75 (#1) · ⚡ 26.6%
  5. " be"803. And to be honest, we�↕ +16.50 (#1) · ⚡ 20.7%
  6. " be". However, to be successful, the project↕ +16.50 (#1) · ⚡ 27.7%
  7. " be" Email↵↵To be honest, there isn↕ +16.38 (#1) · ⚡ 25.6%
  8. " be" following questions. To be on-topic the↕ +16.00 (#1) · ⚡ 28.5%
  9. " be"↵First, to be clear: I know↕ +15.94 (#1) · ⚡ 27.3%
  10. " be" zip code needs to be 5 digits.↵↕ +15.88 (#1) · ⚡ 19.0%
  11. " be" Hebrew Scriptures, to be able to competently↕ +15.81 (#1) · ⚡ 30.1%
  12. " be"<|endoftext|>Want to be tough on the tarn↕ +15.81 (#1) · ⚡ 23.1%
  13. " be" the sentences were to be served concurrently. Over↕ +15.81 (#1) · ⚡ 31.4%
  14. " be" them is going to be deported.↵↵↕ +15.75 (#1) · ⚡ 33.0%
  15. " be" working (and to be frank I've always↕ +15.75 (#1) · ⚡ 24.0%
  16. " be" imperfect history, to be sure, but there↕ +15.75 (#1) · ⚡ 25.6%
  17. " be" system, data to be transmitted are divided for↕ +15.75 (#1) · ⚡ 28.9%
  18. " be" Upon Which Relief Can be Granted and, Alternatively↕ +15.69 (#1) · ⚡ 26.6%
  19. " Be" Footage Can't Be FOIA'ed;↕ +15.62 (#1) · ⚡ 22.1%
  20. " be" years, and to be honest I don't↕ +15.62 (#1) · ⚡ 20.3%
GPT-2 Small · L4/C103

Numerations/lists

Confidence 5/5
Component profile — C103 · layer 4 · negative tail

Logit-lens tokens · negative

#1" etc"#2"etc"#3" assorted"#4"whatever"#5" ect"#6" yes"#7"…)"#8" blah"#9" carbohyd"#10" respectively"

R-lens tokens · negative

#1" etc"#2" Philippine"#3" Filipino"#4" Hispanic"#5" Chuck"#6" Sonny"#7" Manny"#8" Philippines"#9" Furious"#10" nonviolent"

Skewness

-5.53

Excess kurtosis

58.31rank #183

Logcosh deviation

0.1332rank #104

Top-score occurrences · negative

  1. "," the Music, Echo, Volume and PA/↕ -16.62 (#1) · ⚡ 50.0%
  2. "," with eBay, Facebook, etc — but they↕ -16.50 (#1) · ⚡ 45.7%
  3. "," like mood, weight, and health. For↕ -16.50 (#1) · ⚡ 51.6%
  4. "," - friends, family, your company or your↕ -16.25 (#1) · ⚡ 52.7%
  5. "," pork, shellfish, and the eating of↕ -16.00 (#1) · ⚡ 50.4%
  6. ",", cameras, watches, doors opening, to↕ -15.94 (#1) · ⚡ 49.2%
  7. "," a weapon, spell, armor etc usually),↕ -15.88 (#1) · ⚡ 43.2%
  8. "," in education, experience, age, location,↕ -15.81 (#1) · ⚡ 49.0%
  9. "," a homeowner, parent, outdoorsmen, or↕ -15.69 (#1) · ⚡ 37.9%
  10. "," own culture, tradition, and identity. Not↕ -15.56 (#1) · ⚡ 47.9%
  11. ",", actions, claims, proceedings, demands,↕ -15.56 (#1) · ⚡ 36.5%
  12. "," NFL, NBA Live, UFC, The Sims↕ -15.50 (#1) · ⚡ 43.2%
  13. "," – family, friends, society at large –↕ -15.50 (#1) · ⚡ 44.5%
  14. ",", CNN, Forbes, BBC, Guardian).↕ -15.44 (#1) · ⚡ 38.7%
  15. "," as sex, age, Stanford types of A↕ -15.38 (#1) · ⚡ 42.4%
  16. "," of Health, Medicaid, and the Centers for↕ -15.38 (#1) · ⚡ 51.2%
  17. "," of food, shelter, clothing, and health↕ -15.38 (#1) · ⚡ 47.5%
  18. ",", bronchitis, wheezing and↕ -15.31 (#1) · ⚡ 54.3%
  19. "," with food, games, live entertainment, prizes↕ -15.25 (#1) · ⚡ 49.6%
  20. "," workstation, server, router, or switch↕ -15.25 (#1) · ⚡ 43.8%
GPT-2 Small · L5/C66

He

Confidence 5/5
Component profile — C66 · layer 5 · positive tail

Logit-lens tokens · positive

#1"'ll"#2"'d"#3"zbollah"#4"resy"#5"pherd"#6"'ve"#7"eded"#8" knew"#9" encount"#10" wanted"

R-lens tokens · positive

#1" himself"#2" his"#3" he"#4" He"#5"he"#6"He"#7" His"#8" Himself"#9"His"#10" him"

Skewness

+6.33

Excess kurtosis

80.76rank #89

Logcosh deviation

0.1387rank #67

Top-score occurrences · positive

  1. "He" basket.↵↵He was half-right↕ +20.50 (#1) · ⚡ 50.4%
  2. " He" materials”. He then advised the defense↕ +19.88 (#1) · ⚡ 52.7%
  3. " He" of ten lengths. He went on to win↕ +19.62 (#1) · ⚡ 56.2%
  4. "He"↵Early life↵He has a BBA↕ +19.50 (#1) · ⚡ 46.3%
  5. " he" you don't, he'll break your legs↕ +19.38 (#1) · ⚡ 38.7%
  6. "He"”↵↵He said there were fewer↕ +19.38 (#1) · ⚡ 47.1%
  7. " He" an extent that . He thanked those who had↕ +19.38 (#1) · ⚡ 52.7%
  8. "He" conflict.↵↵He added:↵"↕ +19.25 (#1) · ⚡ 49.8%
  9. " he" months of waiting, he walked towards me for↕ +19.00 (#1) · ⚡ 47.7%
  10. " he" relationship,” he said. “↕ +19.00 (#1) · ⚡ 35.4%
  11. " he" record,” he continued.↵↵↕ +18.88 (#1) · ⚡ 37.1%
  12. "He" Teamwork." "He's right." "↕ +18.75 (#1) · ⚡ 44.1%
  13. "He"”↵↵He has his agenda set↕ +18.75 (#1) · ⚡ 54.3%
  14. "He" be."↵↵He said a status hearing↕ +18.62 (#1) · ⚡ 46.5%
  15. "he" Indigenous children only (hepatitis A and↕ +18.62 (#1) · ⚡ 36.1%
  16. " he" connectivity,” he said.↵↵↕ +18.62 (#1) · ⚡ 44.7%
  17. "He" goods.”He continued, “↕ +18.50 (#1) · ⚡ 39.5%
  18. " he"� The show, he says, will examine↕ +18.50 (#1) · ⚡ 35.2%
  19. " he" stalled,” he said in a statement↕ +18.50 (#1) · ⚡ 42.2%
  20. " he" sometimes? Yeah," he wrote. “↕ +18.50 (#2) · ⚡ 23.7%
GPT-2 Small · L5/C312

?

Confidence 0/5
Component profile — C312 · layer 5 · positive tail

Logit-lens tokens · positive

#1"'s"#2" itself"#3" herself"#4" is"#5" alone"#6" himself"#7" was"#8" II"#9","#10" inhibitors"

R-lens tokens · positive

#1":"#2" only"#3" exclusively"#4" observed"#5" DG"#6" successfully"#7" itself"#8" registered"#9" depicted"#10" subscribed"

Skewness

+2.25

Excess kurtosis

12.87rank #530

Logcosh deviation

0.0755rank #313

Top-score occurrences · positive

  1. " MEN" their performance THE HIT MEN share great back stories↕ +13.00 (#1) · ⚡ 15.7%
  2. "LET"utic effects of HAMLET have been demonstrated in↕ +12.69 (#1) · ⚡ 25.4%
  3. "CO" decreased expression of MARCO in intratumoral↕ +11.94 (#1) · ⚡ 23.5%
  4. "Arc" licensing agreements, LifeArc provides funding for academic↕ +11.75 (#1) · ⚡ 13.2%
  5. " MEN"ers, THE HIT MEN, are these actual↕ +11.75 (#1) · ⚡ 11.5%
  6. "ENSE" for HM_SENSE. In vivo imaging↕ +11.75 (#1) · ⚡ 18.9%
  7. "NB"ural planning of PNB, because of its↕ +11.56 (#1) · ⚡ 17.5%
  8. "3"ovital-H3 with sulphonam↕ +11.38 (#1) · ⚡ 19.6%
  9. "LET" cancers, and HAMLET limits the progression of↕ +11.31 (#1) · ⚡ 18.1%
  10. "J" clinical presentation of ONJ and the most recently↕ +11.25 (#1) · ⚡ 21.4%
  11. "OC" technology for UV FSOC has been damp↕ +11.06 (#1) · ⚡ 17.9%
  12. " Eye" the way Tourist Eye will help find restaurants↕ +11.06 (#1) · ⚡ 12.6%
  13. " Terra" funds from Eno Terra will go directly to↕ +11.00 (#1) · ⚡ 12.5%
  14. "itty"ADE 8-Bitty functions as a wireless↕ +10.94 (#1) · ⚡ 14.9%
  15. "aries" Hentai Diaries by writing a review↕ +10.88 (#1) · ⚡ 10.2%
  16. " All" found that Medicare for All would save $450↕ +10.81 (#2) · ⚡ 10.6%
  17. " Shake" up, 070 Shake had plenty of love↕ +10.81 (#1) · ⚡ 11.6%
  18. "J" tools for online ONJ education and this was↕ +10.81 (#1) · ⚡ 20.3%
  19. "ik"k and Chignik during 1940. Meet↕ +10.81 (#1) · ⚡ 12.7%
  20. "SS"-14 or SeSS the arginine↕ +10.75 (#1) · ⚡ 14.9%
GPT-2 Small · L5/C333

Studies

Confidence 5/5
Component profile — C333 · layer 5 · negative tail

Logit-lens tokens · negative

#1" conducted"#2" methodology"#3" findings"#4" published"#5" undertaken"#6" abroad"#7" examining"#8" performed"#9" commissioned"#10" corrobor"

R-lens tokens · negative

#1" studies"#2" study"#3"Study"#4" research"#5" researchers"#6" findings"#7" Study"#8" testing"#9"study"#10" studying"

Skewness

-2.83

Excess kurtosis

32.28rank #288

Logcosh deviation

0.0723rank #334

Top-score occurrences · negative

  1. " studies"]\]. Some studies have examined the association↕ -21.25 (#1) · ⚡ 42.6%
  2. " studies"↵Existing empirical studies often examine people's↕ -19.88 (#1) · ⚡ 43.6%
  3. " study". Because a new study with rats shows that↕ -19.88 (#1) · ⚡ 34.8%
  4. " studies"�↵↵Previous studies only compared female bones↕ -19.75 (#1) · ⚡ 38.1%
  5. " studies" clearly understood. Previous studies have documented a 12↕ -19.62 (#1) · ⚡ 42.8%
  6. " study"↵"A prospective study of 813 bicycle↕ -19.62 (#1) · ⚡ 48.0%
  7. " studies"B26]). Prior studies have demonstrated that t↕ -19.62 (#1) · ⚡ 38.3%
  8. " studies"]^. Previous studies^[@CR↕ -19.50 (#1) · ⚡ 38.1%
  9. " studies".↵Several recent studies identified that postoperative↕ -19.38 (#1) · ⚡ 49.8%
  10. " study"↵↵A new study provides the first evidence↕ -19.38 (#1) · ⚡ 32.2%
  11. " study"According to a 2007 study in the Journal of↕ -19.25 (#1) · ⚡ 38.5%
  12. " study"↵But the new study suggests that this strategy↕ -19.25 (#1) · ⚡ 34.6%
  13. " study", a newly published study has suggested that our↕ -19.12 (#1) · ⚡ 37.1%
  14. " studies"]).↵↵Many studies have suggested ∼40↕ -19.00 (#1) · ⚡ 43.4%
  15. " study" device.↵One study by Pittet et↕ -19.00 (#1) · ⚡ 40.4%
  16. " studies". Additionally, some studies have shown that LIM↕ -18.88 (#1) · ⚡ 40.2%
  17. " studies". Recently, many studies showed that licor↕ -18.75 (#1) · ⚡ 45.3%
  18. " studies"↵A number of studies have investigated the role↕ -18.75 (#1) · ⚡ 37.7%
  19. " study".↵↵A study, published in the↕ -18.62 (#1) · ⚡ 42.6%
  20. " study"↵In a previous study, we identified (-↕ -18.62 (#1) · ⚡ 39.5%
GPT-2 Small · L5/C482

Post-{ newline

Confidence 5/5
Component profile — C482 · layer 5 · negative tail

Logit-lens tokens · negative

#1"}"#2"};"#3" }"#4"}{"#5"})"#6" })"#7"}}"#8"}\""#9" ]"#10" });"

R-lens tokens · negative

#1"}}"#2"}."#3"}"#4"】"#5" }}"#6"};"#7" ]"#8" },"#9" }"#10"}}}"

Skewness

-1.69

Excess kurtosis

12.21rank #542

Logcosh deviation

0.0585rank #483

Top-score occurrences · negative

  1. "\n" {↵ ↕ -11.56 (#1) · ⚡ 23.6%
  2. "\n" {↵ ↕ -11.25 (#1) · ⚡ 23.0%
  3. "\n"s: [{↵ ↕ -11.25 (#1) · ⚡ 19.4%
  4. " "": [{↵ ↕ -11.12 (#1) · ⚡ 25.4%
  5. "\n" {↵ ↕ -11.12 (#1) · ⚡ 19.2%
  6. "\n"↵ {↵ ↕ -11.06 (#1) · ⚡ 20.8%
  7. "\n" {↵ ↕ -11.06 (#1) · ⚡ 21.1%
  8. "\n" {↵ ↕ -11.00 (#1) · ⚡ 20.6%
  9. "\n"↵ return {↵ ↕ -10.94 (#1) · ⚡ 17.4%
  10. "\n" return {↵ ↕ -10.94 (#1) · ⚡ 18.1%
  11. " ": [{↵ ↕ -10.88 (#1) · ⚡ 27.0%
  12. "\n" {↵ ↕ -10.88 (#1) · ⚡ 22.9%
  13. "\n" {↵ ↕ -10.88 (#1) · ⚡ 22.0%
  14. " " {↵ ↕ -10.81 (#1) · ⚡ 27.0%
  15. " "h": {↵ ↕ -10.75 (#1) · ⚡ 23.0%
  16. " " {↵ ↕ -10.75 (#1) · ⚡ 29.9%
  17. " "versions": {↵ "stable↕ -10.69 (#1) · ⚡ 21.8%
  18. " "[↵{↵ "id": 1↕ -10.56 (#1) · ⚡ 20.6%
  19. "\n" {↵ ↕ -10.50 (#1) · ⚡ 20.2%
  20. "U" id_{{UEditor.name}}↕ -10.50 (#2) · ⚡ 11.8%
GPT-2 Small · L5/C598

Pre

Confidence 5/5
Component profile — C598 · layer 5 · negative tail

Logit-lens tokens · negative

#1"requisites"#2"req"#3"loading"#4"anamo"#5"hend"#6"fecture"#7" shipping"#8"emin"#9"nant"#10"empt"

R-lens tokens · negative

#1" Kinder"#2" Pre"#3" PRE"#4" Primal"#5" prep"#6" pre"#7"Pre"#8"prim"#9"pre"#10"req"

Skewness

-1.02

Excess kurtosis

10.85rank #562

Logcosh deviation

0.0440rank #599

Top-score occurrences · negative

  1. " pre"L), Australia's pre-eminent semi↕ -14.75 (#1) · ⚡ 20.2%
  2. " pre" CT has a better preprocedural planning↕ -14.69 (#1) · ⚡ 19.0%
  3. " pre" basic things are already pre-generated. This↕ -14.69 (#1) · ⚡ 21.5%
  4. " pre" confirm that they are prepping another charity LP↕ -14.56 (#1) · ⚡ 19.8%
  5. " pre"↵used as a pre-built set of↕ -14.50 (#1) · ⚡ 21.2%
  6. " pre" almost certain to be pre-conquest,↕ -14.19 (#1) · ⚡ 21.2%
  7. " pre" secret diary of a pre-adolescent daughter↕ -14.12 (#1) · ⚡ 20.0%
  8. " pre" protocols often use a preamble sequence at↕ -14.12 (#1) · ⚡ 22.2%
  9. " pre" deprivation. We performed pre-specified subgroup↕ -14.12 (#1) · ⚡ 20.5%
  10. " pre" reduce the likelihood of preterm delivery, the↕ -14.00 (#1) · ⚡ 21.7%
  11. " pre" Swan Lake in her pre-child past,↕ -14.00 (#1) · ⚡ 19.6%
  12. " pre" with a history of preterm birth. Following↕ -14.00 (#1) · ⚡ 21.3%
  13. " pre" is handled by the preprocessor, so it↕ -13.94 (#1) · ⚡ 20.4%
  14. " pre" juvenile justice facilities: pre- and postrelease↕ -13.88 (#1) · ⚡ 17.8%
  15. " pre" money going through undergraduate prerequisites feels like a↕ -13.88 (#1) · ⚡ 19.7%
  16. " pre" be associated with a pre-operative resistance of↕ -13.88 (#1) · ⚡ 21.7%
  17. " pre". Takes care of pre-screening all↕ -13.88 (#1) · ⚡ 21.4%
  18. " pre" football team in the pre-eliminary squad↕ -13.81 (#1) · ⚡ 18.8%
  19. " pre" included↵↵Page preloader↵↵404↕ -13.81 (#1) · ⚡ 17.7%
  20. " pre" largest city in the pre-Columbian Americas↕ -13.81 (#1) · ⚡ 16.9%
GPT-2 Small · L5/C751

?

Confidence 0/5
Component profile — C751 · layer 5 · negative tail

Logit-lens tokens · negative

#1"‎"#2"wagen"#3" alive"#4"owe"#5".''"#6" Nap"#7"SAN"#8"…."#9" peacefully"#10"wit"

R-lens tokens · negative

#1" Mandela"#2" Boko"#3" Nigerian"#4" awa"#5" Kenyan"#6" African"#7" Negro"#8" gorilla"#9" welf"#10" concess"

Skewness

-0.01

Excess kurtosis

0.91rank #754

Logcosh deviation

0.0126rank #752

Top-score occurrences · negative

  1. "et" be brain-penetrant or -im↕ -8.12 (#2) · ⚡ 4.8%
  2. "et" or cell-penetrating peptides,↕ -7.34 (#2) · ⚡ 4.0%
  3. "et" had thoroughly↵penetrated to his brain↕ -7.12 (#2) · ⚡ 3.6%
  4. "et"istic membrane-penetrating properties have been↕ -6.47 (#2) · ⚡ 3.3%
  5. "st" until OS is reinstalled.↵You↕ -6.41 (#3) · ⚡ 3.7%
  6. "l"hangelsk Oblasts. Air Force↕ -6.16 (#6) · ⚡ 2.7%
  7. " alk" larval mosquito gut alkalinization will be↕ -6.09 (#2) · ⚡ 3.2%
  8. "et" and non-penetrant drugs was drawn↕ -5.91 (#2) · ⚡ 3.6%
  9. "165"638-5165 or email westford↕ -5.81 (#3) · ⚡ 3.3%
  10. "ox"oalkylsiloxanes) in the↕ -5.78 (#3) · ⚡ 4.0%
  11. " =" 3 trials (n = 62 673)↕ -5.75 (#6) · ⚡ 1.8%
  12. "umb" ischemic penumbra, we will↕ -5.72 (#3) · ⚡ 2.8%
  13. "�"�典：Google�↕ -5.69 (#5) · ⚡ 2.7%
  14. "et"↵↵"Monetisation" is pretty↕ -5.66 (#5) · ⚡ 2.1%
  15. "uck" American bare-knuckled boxer. He↕ -5.62 (#1) · ⚡ 2.8%
  16. "れ"�売され、マニア↕ -5.59 (#2) · ⚡ 2.3%
  17. "st"ty↵↵Obstetrics & Gy↕ -5.56 (#5) · ⚡ 1.7%
  18. "st" Manassas Prostitution Attorney↵↵↕ -5.56 (#7) · ⚡ 2.5%
  19. "|" that $|x|\ge 2$↕ -5.53 (#3) · ⚡ 2.6%
  20. "-" the professional-patient-family relationship.↵↕ -5.50 (#3) · ⚡ 3.4%
GPT-2 Small · L6/C159

://

Confidence 5/5
Component profile — C159 · layer 6 · negative tail

Logit-lens tokens · negative

#1"www"#2"://"#3" www"#4"past"#5"docs"#6"natureconservancy"#7":/"#8"youtu"#9"imgur"#10"github"

R-lens tokens · negative

#1"www"#2"://"#3" www"#4"youtu"#5"http"#6" http"#7"github"#8"https"#9" https"#10"cdn"

Skewness

-7.00

Excess kurtosis

156.36rank #5

Logcosh deviation

0.0991rank #160

Top-score occurrences · negative

  1. "://"EN" "https://www.w3↕ -29.38 (#1) · ⚡ 49.8%
  2. "://"↵ // https://tools.ietf↕ -29.12 (#1) · ⚡ 62.9%
  3. "://" ↵// http://www.boost.↕ -29.12 (#1) · ⚡ 56.2%
  4. "://".↵// https://www.jsonr↕ -29.00 (#1) · ⚡ 60.5%
  5. "://"EN" "http://www.w3↕ -29.00 (#1) · ⚡ 50.4%
  6. "://", see <http://www.gnu.↕ -29.00 (#1) · ⚡ 54.3%
  7. "://"EN" "http://www.w3↕ -28.88 (#1) · ⚡ 50.4%
  8. "://" https://www.apache.↕ -28.75 (#1) · ⚡ 48.4%
  9. "://".↵ * https://cs.chromium↕ -28.62 (#1) · ⚡ 60.9%
  10. "://", see <http://www.gnu.↕ -28.62 (#1) · ⚡ 56.2%
  11. "://" signal:↵https://www.bog↕ -28.62 (#1) · ⚡ 60.2%
  12. "://"↵// http://www.apache.↕ -28.50 (#1) · ⚡ 50.4%
  13. "://"url] => https://drive.google.↕ -28.50 (#1) · ⚡ 53.1%
  14. "://"url] => https://drive.google.↕ -28.50 (#1) · ⚡ 59.0%
  15. "://" here:↵https://commons.apache↕ -28.38 (#1) · ⚡ 66.4%
  16. "://"IGN↵// https://stackoverflow.↕ -28.38 (#1) · ⚡ 59.4%
  17. "://", see <http://www.gnu.↕ -28.38 (#1) · ⚡ 56.6%
  18. "://" or copy at http://www.boost.↕ -28.38 (#1) · ⚡ 55.1%
  19. "://" at↵ * https://www.openssl↕ -28.38 (#1) · ⚡ 55.1%
  20. "://", see <http://www.gnu.↕ -28.38 (#1) · ⚡ 57.0%
GPT-2 Small · L6/C174

?

Confidence 0/5
Component profile — C174 · layer 6 · positive tail

Logit-lens tokens · positive

#1"ax"#2"zon"#3"itar"#4"act"#5"inar"#6"os"#7"af"#8"ost"#9"am"#10"lem"

R-lens tokens · positive

#1"escription"#2"ometimes"#3"\u0004"#4"\u0003"#5"�"#6"�"#7"\u0001"#8"\u0002"#9"�"#10"�"

Skewness

+2.83

Excess kurtosis

16.06rank #445

Logcosh deviation

0.0961rank #175

Top-score occurrences · positive

  1. " Fr"The birdlife of Frithy Wood has↕ +11.38 (#1) · ⚡ 12.0%
  2. " war"ogenetics data on warfarin. Therefore↕ +11.38 (#1) · ⚡ 14.4%
  3. " Ft"Defense attorneys wrote that Ftouhi was depressed↕ +11.31 (#1) · ⚡ 10.3%
  4. " Mat", I think whether Mathematica can successfully↕ +11.31 (#1) · ⚡ 15.1%
  5. " su" but we demonstrate that suberin acts as↕ +11.19 (#1) · ⚡ 12.6%
  6. "fire"access_key=firewall_cred↕ +11.12 (#1) · ⚡ 10.2%
  7. " Fre" search, then the Freelancer profiles with↕ +11.06 (#1) · ⚡ 12.6%
  8. " Cord"↵↵To develop Cordova apps for Windows↕ +10.94 (#1) · ⚡ 10.3%
  9. " Lav" what it reveals about Lavoisier's career↕ +10.88 (#1) · ⚡ 10.9%
  10. " loc" survival. Use of locoregional therapies↕ +10.88 (#1) · ⚡ 11.2%
  11. " Je" the biggest misconceptions about Jeju's tourism boom↕ +10.81 (#1) · ⚡ 12.5%
  12. " Ul" volatiles of Ulmo honey were isolated↕ +10.75 (#1) · ⚡ 11.7%
  13. " Ger"↵However you take Gerovital, you↕ +10.75 (#1) · ⚡ 11.1%
  14. " Ib" seven and eight of Ibbi-Sin's↕ +10.75 (#1) · ⚡ 11.9%
  15. " sent", the validity of sentinel node biopsy↕ +10.75 (#1) · ⚡ 13.8%
  16. " ir"okinetic parameters of irinotecan↕ +10.69 (#1) · ⚡ 12.7%
  17. " El" The trial demonstrated that Eloctate was effective↕ +10.69 (#1) · ⚡ 13.9%
  18. " Sent" rate, and enables Sentosa Care to invest↕ +10.69 (#1) · ⚡ 10.5%
  19. " Excel"↵Our motto at Excelsior is �↕ +10.62 (#1) · ⚡ 10.3%
  20. " turn" capacities of the two turnip landraces↕ +10.62 (#1) · ⚡ 12.9%
GPT-2 Small · L7/C215

Combining, Interaction

Confidence 5/5
Component profile — C215 · layer 7 · positive tail

Logit-lens tokens · positive

#1" between"#2" with"#3" Between"#4" among"#5"between"#6" partners"#7" seamlessly"#8" amongst"#9" partner"#10" disparate"

R-lens tokens · positive

#1" between"#2"between"#3" together"#4" mutual"#5" disparate"#6" partners"#7" mutually"#8" partner"#9" shared"#10" alliances"

Skewness

+2.86

Excess kurtosis

22.25rank #335

Logcosh deviation

0.0834rank #216

Top-score occurrences · positive

  1. " merge" and machine consciousness will merge, transcending biological↕ +15.75 (#1) · ⚡ 21.9%
  2. " joint" last week that a joint Jordanian-Israeli command↕ +14.62 (#1) · ⚡ 17.2%
  3. " merged" of Nautical Studies merged in 2010 to establish↕ +14.62 (#1) · ⚡ 18.1%
  4. " merged", that will be merged into the new Berlin↕ +14.50 (#1) · ⚡ 17.1%
  5. " interaction"als. Given the interaction between these genes in↕ +14.38 (#1) · ⚡ 24.8%
  6. "ashing" codons by clashing with Met-t↕ +14.31 (#1) · ⚡ 21.1%
  7. " mix" war that you never mix objectives. Because when↕ +14.19 (#1) · ⚡ 19.0%
  8. " mutual" patients, based on mutual recognition and shared decision↕ +14.12 (#1) · ⚡ 16.8%
  9. " partnership") represents a collaborative partnership leveraging existing educational courses↕ +13.88 (#1) · ⚡ 22.4%
  10. " combined" reports on Thursday, combined large-scale pesticide↕ +13.88 (#1) · ⚡ 16.3%
  11. " mutual"]). The inference of mutual communication (see People↕ +13.81 (#1) · ⚡ 16.6%
  12. " alliance" Federal Reserve Act in alliance with the Rothschild bloc↕ +13.69 (#1) · ⚡ 20.0%
  13. "ges" topmost projection merges with a verbal head↕ +13.62 (#1) · ⚡ 20.4%
  14. " partnership" years of the formal partnership.↵↵The↕ +13.62 (#1) · ⚡ 20.3%
  15. " mutual"itution of discipline and mutual respect in our schools↕ +13.56 (#1) · ⚡ 15.6%
  16. " interaction" the approach envisaging interaction between neurology,↕ +13.50 (#1) · ⚡ 26.2%
  17. " combined" theoretical results have been combined with the experimental data↕ +13.50 (#1) · ⚡ 17.1%
  18. " partnership"This study reflects a partnership between service providers (↕ +13.44 (#1) · ⚡ 22.8%
  19. " joint" unsupervised, joint or sole custody with↕ +13.44 (#1) · ⚡ 14.6%
  20. " interaction" demonstrated the existence of interaction between TSG and↕ +13.38 (#1) · ⚡ 20.4%
GPT-2 Small · L7/C222

Been

Confidence 5/5
Component profile — C222 · layer 7 · positive tail

Logit-lens tokens · positive

#1" lately"#2" since"#3" extensively"#4" recently"#5" steadily"#6" numerous"#7" countless"#8" previously"#9" anecd"#10" remarkably"

R-lens tokens · positive

#1" lately"#2" recently"#3" extensively"#4" since"#5" hitherto"#6" numerous"#7" recent"#8" previously"#9" countless"#10" successfully"

Skewness

+3.26

Excess kurtosis

34.15rank #236

Logcosh deviation

0.0818rank #223

Top-score occurrences · positive

  1. " been"ley says he's been trying to teach his↕ +18.25 (#1) · ⚡ 38.3%
  2. " been"↵Ocon has been left watching Formula 1↕ +17.62 (#1) · ⚡ 34.2%
  3. " been" hearing. He has been better in keeping a↕ +17.62 (#1) · ⚡ 38.9%
  4. " been" NSLP has been widely criticized; however↕ +17.50 (#1) · ⚡ 35.4%
  5. " been"Cousins has been unable to participate in↕ +17.25 (#1) · ⚡ 33.8%
  6. " been"es that she has been in love with Rubin↕ +17.12 (#1) · ⚡ 34.8%
  7. " been" brand and franchise have been in the industry for↕ +17.12 (#1) · ⚡ 38.5%
  8. " been"\]. It has been defined as the ability↕ +17.12 (#1) · ⚡ 32.6%
  9. " been"Mountain lions have been observed in Rancho↕ +17.00 (#1) · ⚡ 33.4%
  10. " been"<|endoftext|>I've been successfully using XAM↕ +17.00 (#1) · ⚡ 32.2%
  11. " been", Pennsylvania, has been processing postconsumer food↕ +16.88 (#1) · ⚡ 34.4%
  12. " been" Her friend that has been skinny for her whole↕ +16.88 (#1) · ⚡ 34.0%
  13. " been" Awards. Miranda has been nominated for an Oscar↕ +16.88 (#1) · ⚡ 34.4%
  14. " been"}$$ which has been studied in detail by↕ +16.88 (#1) · ⚡ 34.4%
  15. " been" I’ve been working on. I↕ +16.88 (#1) · ⚡ 28.3%
  16. " been" a person who's been voting if he wants↕ +16.88 (#1) · ⚡ 32.0%
  17. " been" treatment. He has been detained by police and↕ +16.75 (#1) · ⚡ 32.6%
  18. " been" to an abortion has been the law since 1973↕ +16.75 (#1) · ⚡ 30.1%
  19. " been". The applicant has been trained in somatic↕ +16.75 (#1) · ⚡ 37.1%
  20. " been"Well, I've been waiting patiently over a↕ +16.75 (#1) · ⚡ 33.4%
GPT-2 Small · L7/C347

Name

Confidence 5/5
Component profile — C347 · layer 7 · positive tail

Logit-lens tokens · positive

#1"plates"#2" names"#3"names"#4"plate"#5"paces"#6"name"#7" name"#8" aliases"#9" naming"#10"prefix"

R-lens tokens · positive

#1" surname"#2" name"#3" names"#4" naming"#5" Names"#6"names"#7" surn"#8"Names"#9"name"#10" initials"

Skewness

+2.42

Excess kurtosis

26.41rank #291

Logcosh deviation

0.0650rank #348

Top-score occurrences · positive

  1. " name"eline↵↵The name Adeline has been↕ +21.88 (#1) · ⚡ 40.0%
  2. " name"Bahri. The name refers to hypost↕ +20.25 (#1) · ⚡ 42.0%
  3. " name"↵↵As the name suggests, the book↕ +20.00 (#1) · ⚡ 22.9%
  4. " name" The abbreviated trading name ACC was adopted in↕ +19.88 (#1) · ⚡ 35.2%
  5. " name"↵The robot's name is "Patrick"↕ +19.88 (#1) · ⚡ 35.2%
  6. " name" repository. Despite the name, these packages will↕ +19.62 (#1) · ⚡ 32.8%
  7. " name", hence their previous name Northern Bombers)↕ +19.50 (#1) · ⚡ 30.9%
  8. " name"Final note: The name Badge refers to how↕ +19.50 (#1) · ⚡ 45.3%
  9. " name"lsberg is the name of a district of↕ +19.00 (#1) · ⚡ 34.4%
  10. " name" spring, hence the name.↵↵See↕ +19.00 (#1) · ⚡ 27.0%
  11. " name" the Māori name of the range protected↕ +18.88 (#1) · ⚡ 38.9%
  12. " name"↵Is there a name for the center of↕ +18.62 (#1) · ⚡ 34.4%
  13. " name"oted to change its name to the Rutledge↕ +18.38 (#1) · ⚡ 30.7%
  14. " name" (hence the name, Perfectly Frank↕ +18.38 (#1) · ⚡ 29.1%
  15. " name"ational and changed its name to The White Mountain↕ +18.38 (#1) · ⚡ 29.3%
  16. " name", and underwent several name changes until its current↕ +18.38 (#1) · ⚡ 37.5%
  17. " name"aque" (product name) of the East↕ +18.38 (#1) · ⚡ 35.4%
  18. " name"The anglicised name Letterbreckaun↕ +18.25 (#1) · ⚡ 42.8%
  19. " name" right back. His name is Olly.↕ +18.25 (#1) · ⚡ 28.5%
  20. " name": four. His name, my name.↕ +18.25 (#1) · ⚡ 31.8%
GPT-2 Small · L7/C380

Post-colon newline

Confidence 5/5
Component profile — C380 · layer 7 · negative tail

Logit-lens tokens · negative

#1"<?"#2"https"#3"Quote"#4"http"#5"pmwiki"#6"Copy"#7"``"#8"sudo"#9"Example"#10"$$$$"

R-lens tokens · negative

#1"<?"#2"\n"#3"https"#4" https"#5"#"#6">>>"#7"    "#8"$$"#9" →"#10"<"

Skewness

-2.56

Excess kurtosis

25.65rank #293

Logcosh deviation

0.0617rank #381

Top-score occurrences · negative

  1. "\n"yml`:↵↵```y↕ -16.88 (#1) · ⚡ 47.1%
  2. "\n"↵For example:↵foo.cpp:↕ -16.75 (#1) · ⚡ 47.9%
  3. "\n" like this instead:↵def lol(a↕ -16.75 (#1) · ⚡ 49.6%
  4. "\n" looks like this:↵↵https://api↕ -16.62 (#1) · ⚡ 46.1%
  5. "\n"ivy language:↵<MainToggle↕ -16.38 (#1) · ⚡ 51.2%
  6. "\n" short_name:↵it "should be↕ -16.25 (#1) · ⚡ 47.1%
  7. "\n" would look like:↵package com.example↕ -16.12 (#1) · ⚡ 51.6%
  8. "\n"htaccess file:↵RewriteEngine On↕ -16.12 (#1) · ⚡ 48.4%
  9. "\n"..↵eg:↵el:↵ ↕ -16.00 (#1) · ⚡ 51.6%
  10. "\n" an object file:↵$ objcopy -↕ -15.88 (#1) · ⚡ 42.6%
  11. "\n" ActiveRecord query:↵Order.order("↕ -15.81 (#1) · ⚡ 51.6%
  12. "\n") would look like↵.globl ↕ -15.75 (#1) · ⚡ 39.5%
  13. "\n" the same thing:↵For Each row in↕ -15.69 (#1) · ⚡ 51.6%
  14. "\n" could just run:↵$obj = Get↕ -15.69 (#1) · ⚡ 35.9%
  15. "\n" for this object:↵compare function for↕ -15.62 (#1) · ⚡ 53.9%
  16. "\n" quantity like so:↵$customer->↕ -15.62 (#1) · ⚡ 45.5%
  17. "\n"↵Like this:↵$previousFile↕ -15.62 (#1) · ⚡ 46.9%
  18. "\n" of the relation:↵user = User.↕ -15.62 (#1) · ⚡ 47.9%
  19. "\n" look like this:↵{if is_↕ -15.62 (#1) · ⚡ 45.9%
  20. "\n" ConversationsController:↵class ConversationsController↕ -15.56 (#1) · ⚡ 44.7%
GPT-2 Small · L7/C395

Sentence-ending period

Confidence 5/5
Component profile — C395 · layer 7 · negative tail

Logit-lens tokens · negative

#1" Especially"#2" But"#3" Maybe"#4" And"#5" Everyone"#6" It"#7" If"#8" There"#9" Plenty"#10" Nobody"

R-lens tokens · negative

#1" Everyone"#2" But"#3" Nobody"#4" Maybe"#5" And"#6" Especially"#7" Every"#8" Sure"#9" It"#10" I"

Skewness

-1.90

Excess kurtosis

13.48rank #491

Logcosh deviation

0.0606rank #396

Top-score occurrences · negative

  1. "." not about holding on.”↵↵↕ -13.69 (#1) · ⚡ 26.4%
  2. "."’t matter. All that matters is↕ -12.94 (#1) · ⚡ 33.8%
  3. "." why he wrote it. Becoming a parent↕ -12.94 (#1) · ⚡ 27.5%
  4. "." man for the job. Articulate?↕ -12.94 (#1) · ⚡ 33.0%
  5. "." different as you think. People play so much↕ -12.81 (#1) · ⚡ 29.3%
  6. "."It was a challenge. I lived with it↕ -12.62 (#1) · ⚡ 28.5%
  7. "." we didn't lose. We gave it a↕ -12.50 (#1) · ⚡ 27.7%
  8. "."�t freak out. Rule number one is↕ -12.50 (#1) · ⚡ 22.1%
  9. "." liberals get their way. The policies do the↕ -12.31 (#1) · ⚡ 38.9%
  10. "."You live and learn. Miami is just another↕ -12.31 (#1) · ⚡ 25.6%
  11. "."↵Not a problem.↵Is there a↕ -12.31 (#1) · ⚡ 26.0%
  12. "." you special, then. This Sunday you↵↕ -12.25 (#1) · ⚡ 32.8%
  13. "." I got to leave. I don't know↕ -12.25 (#1) · ⚡ 22.8%
  14. "."’s normal. It’s↕ -12.25 (#1) · ⚡ 36.7%
  15. "." I want more time. I need more time↕ -12.19 (#1) · ⚡ 27.7%
  16. "." I want to pitch. I don’↕ -12.19 (#1) · ⚡ 25.4%
  17. "." seem to do it. Is it because I↕ -12.12 (#1) · ⚡ 29.7%
  18. "."s↵not you. It is not the↕ -12.12 (#1) · ⚡ 30.7%
  19. "." reasons to be happy.↵↵“↕ -12.06 (#1) · ⚡ 28.7%
  20. "." more than a club. They are an institution↕ -12.06 (#1) · ⚡ 15.6%
GPT-2 Small · L7/C422

Repeated variable in later polynomial term

Confidence 5/5
Component profile — C422 · layer 7 · negative tail

Logit-lens tokens · negative

#1"**"#2"^"#3"*"#4" �"#5"##"#6"′"#7"²"#8"***"#9" +"#10"*,"

R-lens tokens · negative

#1"**"#2"^"#3"***"#4"*"#5"##"#6"̶"#7"♦"#8"═"#9"†"#10"*,"

Skewness

-1.92

Excess kurtosis

21.70rank #338

Logcosh deviation

0.0579rank #423

Top-score occurrences · negative

  1. "p"3 - 7*p**2 + 0↕ -17.62 (#1) · ⚡ 38.5%
  2. "m"150 - 3*m**2. Factor↕ -17.25 (#1) · ⚡ 35.4%
  3. "v"5 - 19*v wrt v.↕ -17.12 (#1) · ⚡ 35.9%
  4. "o" + 1657*o**4 - 9↕ -17.12 (#1) · ⚡ 37.1%
  5. "g"4 + 5*g - 7. Let↕ -17.00 (#1) · ⚡ 31.8%
  6. "m"2 + 21*m**2 + 21↕ -17.00 (#1) · ⚡ 33.8%
  7. "r"2 - 64*r**2 wrt↕ -17.00 (#1) · ⚡ 40.0%
  8. "t"2 - 4*t - 115. Let↕ -16.88 (#1) · ⚡ 37.5%
  9. "v"3 + 198*v**3.↵↕ -16.88 (#1) · ⚡ 32.2%
  10. "p"3 + 6*p**2 - p↕ -16.88 (#1) · ⚡ 37.9%
  11. "t" 57 - 3*t**3 + t↕ -16.75 (#1) · ⚡ 35.4%
  12. "t"6 - 6*t**2. Let↕ -16.75 (#1) · ⚡ 39.8%
  13. "f"3 - 21*f**2. Deter↕ -16.75 (#1) · ⚡ 35.2%
  14. "c"2 + 4*c.↵2*↕ -16.75 (#1) · ⚡ 28.7%
  15. "t"6 + 3*t**2/2↕ -16.75 (#1) · ⚡ 35.7%
  16. "m"5 + 11*m**2. Calcul↕ -16.62 (#1) · ⚡ 36.5%
  17. "b")*(4*b - 2*b↕ -16.62 (#1) · ⚡ 40.6%
  18. " p"(p) = p**3 + 10↕ -16.62 (#1) · ⚡ 25.8%
  19. "x" - 493*x**3 + 162↕ -16.62 (#1) · ⚡ 32.0%
  20. "h"3 + 78*h**2 - 10↕ -16.50 (#1) · ⚡ 31.2%
GPT-2 Small · L7/C452

?

Confidence 0/5
Component profile — C452 · layer 7 · negative tail

Logit-lens tokens · negative

#1" is"#2"′"#3" refers"#4"ō"#5"etooth"#6"rite"#7"igation"#8" consists"#9" specializes"#10"recated"

R-lens tokens · negative

#1"recated"#2"ricular"#3"aditional"#4"atchewan"#5"�"#6"trak"#7"ocene"#8"soDeliveryDate"#9" Cosponsors"#10"ategory"

Skewness

-1.44

Excess kurtosis

7.63rank #613

Logcosh deviation

0.0558rank #453

Top-score occurrences · negative

  1. "ay"↵↵—Clay Siegall, the↕ -11.62 (#1) · ⚡ 12.5%
  2. "split" "patterns/split";↵↵/*↕ -11.56 (#1) · ⚡ 9.7%
  3. "an"an Crowe Kieran Crowe is a↕ -11.56 (#1) · ⚡ 10.1%
  4. " Raymond" is relatively large, Raymond mill is developed specifically↕ -11.44 (#1) · ⚡ 11.6%
  5. " Anna" in South Australia. Anna Creek oversees 6,↕ -11.25 (#1) · ⚡ 12.0%
  6. " n" + s and give n.↵267↵↕ -11.19 (#2) · ⚡ 12.3%
  7. " Triangle" here 🡺 Triangle Cash↵↵#↕ -11.00 (#1) · ⚡ 11.5%
  8. " y" + 8. Give y(14).↵↕ -10.69 (#2) · ⚡ 10.8%
  9. "us"l↵↵Magnus Schriefl belongs↕ -10.69 (#1) · ⚡ 8.4%
  10. " Eagle" complete game objectives? Eagle Flight is a bit↕ -10.69 (#1) · ⚡ 9.2%
  11. "A", 0x41A0 ↕ -10.62 (#1) · ⚡ 9.4%
  12. " Kris" Trifari, Kris will be hosting a↕ -10.62 (#1) · ⚡ 8.3%
  13. " Deep"arwals. Deepak convinced Sanji↕ -10.62 (#2) · ⚡ 8.2%
  14. " Roh" Sydney" Dr Peter Rohde is an Australian↕ -10.25 (#2) · ⚡ 8.3%
  15. " h" 24. Determine h(0).↵↕ -10.19 (#3) · ⚡ 10.0%
  16. "Lex"function (CssLexerMode) {↕ -10.06 (#1) · ⚡ 7.9%
  17. "and" the crime. Anand and Tilak had↕ -10.06 (#1) · ⚡ 8.2%
  18. " Bryan" contributor Bryan3000. Bryan and I have been↕ -10.00 (#1) · ⚡ 8.2%
  19. " f"(p). Factor f(h).↵↕ -10.00 (#2) · ⚡ 10.9%
  20. " n"**2. Factor n(t).↵↕ -10.00 (#3) · ⚡ 11.0%
GPT-2 Small · L7/C467

Repeated math questions

Confidence 1/5
Component profile — C467 · layer 7 · positive tail

Logit-lens tokens · positive

#1" regression"#2" enum"#3"olean"#4" finally"#5" quantity"#6" final"#7"ary"#8" spect"#9" boolean"#10" isEnabled"

R-lens tokens · positive

#1" total"#2" size"#3" quantity"#4" the"#5" a"#6" number"#7" name"#8" pass"#9" class"#10" count"

Skewness

+1.42

Excess kurtosis

10.62rank #545

Logcosh deviation

0.0547rank #468

Top-score occurrences · positive

  1. " the" 9. What are the prime factors of a↕ +13.31 (#1) · ⚡ 16.3%
  2. " the"12). What are the prime factors of p↕ +13.06 (#1) · ⚡ 15.6%
  3. "Collect"o**3↵Collect the terms in -↕ +12.62 (#1) · ⚡ 18.6%
  4. " the"19↵What are the prime factors of 4↕ +12.62 (#1) · ⚡ 21.5%
  5. "Collect"n**3↵Collect the terms in -↕ +12.56 (#1) · ⚡ 18.8%
  6. "Collect"j**3↵Collect the terms in -↕ +12.50 (#1) · ⚡ 18.9%
  7. " the" 229↵What are the prime factors of 147↕ +12.44 (#1) · ⚡ 18.8%
  8. " is"*m. What is the tens digit of↕ +12.44 (#1) · ⚡ 15.0%
  9. " is"*s. What is the remainder when z↕ +12.38 (#1) · ⚡ 14.5%
  10. "Collect"3 - 1↵Collect the terms in 5↕ +12.19 (#1) · ⚡ 18.7%
  11. " is": 12}. What is prob of sequence d↕ +11.94 (#1) · ⚡ 11.3%
  12. "Collect"71*w↵Collect the terms in 15↕ +11.81 (#1) · ⚡ 17.3%
  13. " is"↵6↵What is the hundreds digit of↕ +11.81 (#1) · ⚡ 17.1%
  14. " units"↵What is the units digit of 769↕ +11.81 (#1) · ⚡ 10.9%
  15. " List" 12 - w. List the prime factors of↕ +11.75 (#1) · ⚡ 13.0%
  16. " the" + 64. List the prime factors of t↕ +11.75 (#1) · ⚡ 13.7%
  17. "Collect"a**3↵Collect the terms in 8↕ +11.75 (#1) · ⚡ 16.7%
  18. "In"-405233↵In base 2, what↕ +11.75 (#1) · ⚡ 16.5%
  19. " List"2*y. List the prime factors of↕ +11.75 (#1) · ⚡ 14.1%
  20. " is"↵2↵What is the units digit of↕ +11.69 (#1) · ⚡ 17.0%
GPT-2 Small · L7/C625

//

Confidence 5/5
Component profile — C625 · layer 7 · negative tail

Logit-lens tokens · negative

#1"////////////////////////////////"#2"////////////////"#3" prints"#4"////////"#5"Same"#6" Alternatively"#7" alternatively"#8" Oops"#9" NOTE"#10" TOD"

R-lens tokens · negative

#1"////"#2" prints"#3"////////////////////////////////"#4" toget"#5"////////////////"#6" //"#7" alternatively"#8" Print"#9" tiss"#10" Ass"

Skewness

-1.65

Excess kurtosis

23.46rank #311

Logcosh deviation

0.0380rank #626

Top-score occurrences · negative

  1. " //"); //<-doesn't↕ -20.38 (#1) · ⚡ 51.2%
  2. " //" // eslint-↕ -19.50 (#1) · ⚡ 44.1%
  3. " //" = false↵↵ // cycle remove sets for↕ -19.38 (#1) · ⚡ 47.1%
  4. " //"i, 1); // remove groupItem from↕ -19.38 (#1) · ⚡ 43.8%
  5. " //" total = 1; // Initialize the total↕ -19.25 (#1) · ⚡ 42.4%
  6. " //"↵{↵ //do something↵}↕ -19.25 (#1) · ⚡ 40.2%
  7. " //" // TODO Auto-↕ -19.25 (#1) · ⚡ 23.1%
  8. " //"↵ // Explicitly take the↕ -19.12 (#1) · ⚡ 39.8%
  9. " //", 'r'); // open file↵$↕ -19.12 (#1) · ⚡ 41.2%
  10. " //" // set ResponseHandler to↕ -19.12 (#1) · ⚡ 46.3%
  11. " //" \"test\"" // you probably have one↕ -19.12 (#1) · ⚡ 43.9%
  12. " //"↵{↵ //Create object from converter↕ -19.12 (#1) · ⚡ 40.0%
  13. " //" // do something!↵↕ -19.00 (#1) · ⚡ 45.5%
  14. " //"↵{↵ // Identity precondition↕ -19.00 (#1) · ⚡ 39.5%
  15. " //"error(err)); // promise↵↵if↕ -19.00 (#1) · ⚡ 43.8%
  16. " //" != 1) { // If the_cookie↕ -19.00 (#1) · ⚡ 40.6%
  17. " //" null;↵↵ //here goes a function↕ -19.00 (#1) · ⚡ 46.1%
  18. " //" // nothing has changed↵↕ -19.00 (#1) · ⚡ 40.8%
  19. " //" //...↵ ↕ -19.00 (#1) · ⚡ 42.4%
  20. " //" // ignore all properties that↕ -18.88 (#1) · ⚡ 41.8%
GPT-2 Small · L7/C628

Prefix non-

Confidence 5/5
Component profile — C628 · layer 7 · positive tail

Logit-lens tokens · positive

#1"profit"#2"interest"#3"union"#4"yne"#5"erness"#6"compliance"#7"binding"#8"existent"#9"binary"#10"stop"

R-lens tokens · positive

#1" suspic"#2"violence"#3"erness"#4" advoc"#5"independence"#6"union"#7"yne"#8" experien"#9"violent"#10" conclud"

Skewness

+1.42

Excess kurtosis

17.89rank #401

Logcosh deviation

0.0375rank #629

Top-score occurrences · positive

  1. "-"<|endoftext|>Non-equilibrium x-↕ +17.00 (#1) · ⚡ 29.9%
  2. "-"<|endoftext|>Non-linear relationships between the↕ +17.00 (#1) · ⚡ 29.7%
  3. "-"↵↵The non-profit organization launched on↕ +16.75 (#1) · ⚡ 30.3%
  4. "Non" F (Private↵Non Medical Institutions)↕ +16.62 (#1) · ⚡ 24.3%
  5. "-" In addition, non-invasive mechanical ventilation↕ +16.25 (#1) · ⚡ 31.2%
  6. "-", while the non-intercepted portion↕ +16.12 (#1) · ⚡ 27.5%
  7. "-"<|endoftext|>Various non-informational,↕ +15.94 (#1) · ⚡ 28.9%
  8. "-" could help identify non-citizens on the voter↕ +15.88 (#1) · ⚡ 27.0%
  9. "-"affairs and non-theatrical businesses↕ +15.88 (#1) · ⚡ 27.5%
  10. "-"), 30 were non-responders (NR↕ +15.81 (#1) · ⚡ 26.4%
  11. " non". In addition, non-invasive mechanical↕ +15.81 (#1) · ⚡ 28.3%
  12. "-", recruiting a non-European managing director,↕ +15.75 (#1) · ⚡ 28.3%
  13. " non" representing the civilian, non-institutionalized↕ +15.75 (#1) · ⚡ 21.4%
  14. " non", higher levels of non-marital co↕ +15.75 (#1) · ⚡ 28.9%
  15. "-" and has worked non-traditional hours to facilitate↕ +15.69 (#1) · ⚡ 26.0%
  16. "-" editorial material, non-English literature and repeated↕ +15.69 (#1) · ⚡ 29.5%
  17. " non" $1,500 non-repayable↕ +15.62 (#1) · ⚡ 27.0%
  18. " non" finances both DOE and non-DOE R↕ +15.62 (#1) · ⚡ 25.2%
  19. "Non" for Garnishment (NonPeriodic)↕ +15.62 (#1) · ⚡ 24.3%
  20. "-" Mee's non-stop hiccups↕ +15.62 (#1) · ⚡ 23.8%
GPT-2 Small · L7/C757

Post-URL for details or name a few

Confidence 1/5
Component profile — C757 · layer 7 · positive tail

Logit-lens tokens · positive

#1"xit"#2"icter"#3"FK"#4"��"#5" similarly"#6"alog"#7" answ"#8" cures"#9"qual"#10"ciation"

R-lens tokens · positive

#1"????"#2"????????"#3"??"#4"ã"#5"apest"#6"iots"#7"xit"#8"ciation"#9"aghd"#10"alog"

Skewness

+0.07

Excess kurtosis

0.75rank #759

Logcosh deviation

0.0093rank #758

Top-score occurrences · positive

  1. " for"/xsl/ for copyright↵ ↕ +9.69 (#3) · ⚡ 6.8%
  2. " for"44/4/ for the full example:↕ +8.50 (#3) · ⚡ 6.7%
  3. " for"acQ2> for details).↵↵↕ +8.44 (#3) · ⚡ 6.3%
  4. " name" and Mitt Romney to name a few are all↕ +8.06 (#2) · ⚡ 3.4%
  5. " name" overseas customers, to name a few.↵↕ +7.88 (#2) · ⚡ 3.0%
  6. " for"js#readme for more supported runtimes↕ +7.81 (#3) · ⚡ 5.8%
  7. " for" in Amplifiers, for more information.↵↕ +7.62 (#3) · ⚡ 5.4%
  8. " for"Hanson2010] for a review).↵↕ +7.59 (#3) · ⚡ 5.8%
  9. " for"ICENSE.txt for license information↵//↕ +7.44 (#3) · ⚡ 4.2%
  10. " for"ICENSE.txt for license information↵//↕ +7.44 (#3) · ⚡ 4.1%
  11. " among" Paul and Ted Cruz among his political idols.↕ +7.31 (#1) · ⚡ 3.6%
  12. " for" the AUTHORS file for names of contributors.↕ +7.25 (#3) · ⚡ 4.3%
  13. " for"ire/Durandal for details.↵ */↕ +7.25 (#3) · ⚡ 4.7%
  14. " for" in the project root for license information.↵↕ +7.09 (#3) · ⚡ 4.8%
  15. " for" javadoc for QueuedThreadPool↕ +6.94 (#3) · ⚡ 5.6%
  16. " for"(See code below for details),issue is↕ +6.88 (#3) · ⚡ 5.5%
  17. " to"3::Registry to make those locations resolve↕ +6.72 (#3) · ⚡ 5.4%
  18. " now" some DVD players are now region-free.↕ +6.72 (#2) · ⚡ 5.4%
  19. " may" the Kremlin last week may look aggressive toward NATO↕ +6.62 (#3) · ⚡ 4.6%
  20. " for"ENSE.TXT for details.↵//↕ +6.59 (#3) · ⚡ 3.8%
GPT-2 Small · L8/C152

Post-quote reporting verb

Confidence 5/5
Component profile — C152 · layer 8 · negative tail

Logit-lens tokens · negative

#1" sarcast"#2" bluntly"#3" quoted"#4" adding"#5" stressing"#6" rhet"#7" spokesman"#8" noting"#9" referring"#10" stressed"

R-lens tokens · negative

#1" adding"#2" stressing"#3" explaining"#4" \""#5" stressed"#6" noting"#7" citing"#8" spokesman"#9" quoted"#10" elaborated"

Skewness

-4.34

Excess kurtosis

52.80rank #123

Logcosh deviation

0.0959rank #153

Top-score occurrences · negative

  1. " said" buffer copies," Smith said. "Howard thinks↕ -22.62 (#1) · ⚡ 54.3%
  2. " said" Police Chief Chuck Harmon said. "She was↕ -21.50 (#1) · ⚡ 45.3%
  3. " said" to make," he said. Colford explained↕ -21.25 (#1) · ⚡ 49.6%
  4. " said" Bill O'Brien said. "Had some↕ -21.25 (#1) · ⚡ 37.1%
  5. " said" diminishing returns," McCarthy said. "Their pots↕ -21.12 (#1) · ⚡ 46.7%
  6. " said" brains," Woolley said.↵↵W↕ -21.00 (#1) · ⚡ 41.4%
  7. " said" an Ofcom spokesman said.↵↵In↕ -20.88 (#1) · ⚡ 42.6%
  8. " said" of the study, said in a press release↕ -20.88 (#1) · ⚡ 44.5%
  9. " said"," Ellington said. "This is↕ -20.88 (#1) · ⚡ 38.9%
  10. " said" so positive," she said. "I call↕ -20.88 (#1) · ⚡ 40.4%
  11. " said" of justice," Dudley said. "It's↕ -20.75 (#1) · ⚡ 48.0%
  12. " continued" collaborative computing," he continued.↵↵It↕ -20.75 (#1) · ⚡ 36.5%
  13. " said" national effort," he said.↵↵"↕ -20.62 (#1) · ⚡ 48.8%
  14. " said" too small," Patel said. "It's↕ -20.62 (#1) · ⚡ 50.0%
  15. " said" been lower," Lodge said at last Thursday's↕ -20.50 (#1) · ⚡ 48.4%
  16. " said" three months," he said. "The beginning↕ -20.38 (#1) · ⚡ 42.0%
  17. " said"," Dipietro said in his statement.↕ -20.38 (#1) · ⚡ 50.4%
  18. " said" him," Metz said.↵↵Met↕ -20.25 (#1) · ⚡ 42.6%
  19. " said" Mr Cairns said.↵↵"↕ -20.25 (#1) · ⚡ 44.9%
  20. " said" play well," he said. "If I↕ -20.25 (#1) · ⚡ 40.2%
GPT-2 Small · L8/C396

Programming/API specification language

Confidence 5/5
Component profile — C396 · layer 8 · negative tail

Logit-lens tokens · negative

#1" JSON"#2" syntax"#3" initialization"#4" integers"#5" parameter"#6" tuple"#7" constructor"#8" parsed"#9" dynamically"#10" params"

R-lens tokens · negative

#1" JSON"#2" syntax"#3" tuple"#4" parsing"#5" Python"#6" initialization"#7" regex"#8" parser"#9" parsed"#10" JavaScript"

Skewness

-1.84

Excess kurtosis

7.33rank #618

Logcosh deviation

0.0588rank #397

Top-score occurrences · negative

  1. " returns" in Java interface that returns class or subclass type↕ -9.88 (#1) · ⚡ 12.1%
  2. " in" require all modules ending in ".spec" from↕ -9.69 (#1) · ⚡ 10.5%
  3. " an"ategories" is not an array, but object↕ -9.50 (#2) · ⚡ 9.7%
  4. " with" methods whose name starts with test_.↵↵↕ -9.44 (#1) · ⚡ 9.7%
  5. " be" of item. Must be valid after update()↕ -9.44 (#1) · ⚡ 10.6%
  6. " An" [callback] - An optional callback to use↕ -9.44 (#2) · ⚡ 9.0%
  7. " returning" node query that is returning results. However,↕ -9.25 (#1) · ⚡ 12.0%
  8. " with" an identifier cannot start with an underscore. So↕ -9.00 (#1) · ⚡ 10.6%
  9. " not"categories" is not an array, but↕ -8.88 (#2) · ⚡ 10.4%
  10. " object" if they return an object).↵↵A↕ -8.88 (#1) · ⚡ 10.2%
  11. " declared"table should better be declared as a sysname↕ -8.69 (#1) · ⚡ 10.2%
  12. " plain" mentioned actions must be plain objects but when you↕ -8.69 (#1) · ⚡ 6.6%
  13. " pass" an option you can pass to NewConn to↕ -8.69 (#1) · ⚡ 8.2%
  14. " be".↵↵Must be the ARN of↕ -8.69 (#2) · ⚡ 9.5%
  15. " an" do not start with an underscore. So↕ -8.69 (#2) · ⚡ 6.7%
  16. " object" An object containing either "data↕ -8.62 (#1) · ⚡ 9.8%
  17. "Returns"ains ↵↵Returns true if and only↕ -8.56 (#1) · ⚡ 9.5%
  18. " be" error mentioned actions must be plain objects but when↕ -8.56 (#1) · ⚡ 8.4%
  19. " be" 'action' MUST be the same as PHP↕ -8.56 (#1) · ⚡ 10.2%
  20. " object"Model stores the argument object in this.state↕ -8.56 (#1) · ⚡ 11.4%
GPT-2 Small · L8/C743

Mr

Confidence 5/5
Component profile — C743 · layer 8 · positive tail

Logit-lens tokens · positive

#1"anyahu"#2" MOR"#3" Brown"#4"arnaev"#5"Bay"#6" MOV"#7"iann"#8" ILCS"#9" Rain"#10"ifiers"

R-lens tokens · positive

#1"imes"#2" Improvement"#3" Office"#4" ADD"#5" Bloom"#6" highlights"#7" multiplication"#8" evaluations"#9" experiences"#10" annex"

Skewness

+0.24

Excess kurtosis

2.92rank #715

Logcosh deviation

0.0146rank #744

Top-score occurrences · positive

  1. " Mr" about corporate Canada. Mr. Carney, as↕ +13.31 (#2) · ⚡ 14.4%
  2. " Mr" should be doing," Mr. Dayton said of↕ +13.19 (#2) · ⚡ 12.2%
  3. "Mr" Thursday.↵↵Mr Wright started a relationship↕ +13.06 (#2) · ⚡ 11.4%
  4. " Mr"<|endoftext|>But Democrats hope Mr. Bush’↕ +13.06 (#2) · ⚡ 12.3%
  5. "Mr" families.↵↵Mr. McCullagh said↕ +12.94 (#2) · ⚡ 12.7%
  6. " Mr" to reporters present, Mr. O’↕ +12.81 (#2) · ⚡ 12.7%
  7. " Mr" in early January, Mr Stefanini’↕ +12.75 (#1) · ⚡ 15.3%
  8. " Mr" out the sneakers when Mr. Kaepernick told the↕ +12.75 (#2) · ⚡ 13.1%
  9. "Mr" Rosenthal.↵↵Mr. Rosenthal is a↕ +12.44 (#2) · ⚡ 13.8%
  10. " Mr" Pierre Nolin. Mr. Nolin was↕ +12.38 (#1) · ⚡ 13.9%
  11. "Mr" said.↵↵Mr Obama contrasted the reality↕ +12.38 (#1) · ⚡ 12.0%
  12. "Mr" case.↵↵Mr Crosland has written↕ +12.25 (#1) · ⚡ 12.9%
  13. " Mr",” said Mr Ghobril.↕ +12.19 (#1) · ⚡ 13.0%
  14. " Mr" and top management, Mr. Wan Jie↕ +12.00 (#2) · ⚡ 11.2%
  15. " Mr" late.Prosecutors said Mr. Bates has a↕ +12.00 (#2) · ⚡ 13.5%
  16. "Mr" site.↵↵Mr Huljich↕ +12.00 (#1) · ⚡ 13.0%
  17. "Mr" emerged.↵↵Mr Finnie, who↕ +12.00 (#1) · ⚡ 11.4%
  18. " Mr" the operations side, Mr. Quigley↕ +11.94 (#2) · ⚡ 13.3%
  19. " Mr",” says Mr Rogan.↵↕ +11.88 (#1) · ⚡ 13.2%
  20. " Mr" are dead, explains Mr Bernal.↵↕ +11.75 (#2) · ⚡ 11.6%
GPT-2 Small · L9/C56

Do/does

Confidence 5/5
Component profile — C56 · layer 9 · negative tail

Logit-lens tokens · negative

#1" nothing"#2"omsday"#3" wonders"#4" not"#5"pez"#6" things"#7"vet"#8"oms"#9" justice"#10" everything"

R-lens tokens · negative

#1" nothing"#2" things"#3"pez"#4" wonders"#5" NOT"#6" not"#7" something"#8" justice"#9" everything"#10"omsday"

Skewness

-7.09

Excess kurtosis

122.07rank #15

Logcosh deviation

0.1206rank #57

Top-score occurrences · negative

  1. " does" or rigid tubing which does not require severing↕ -24.38 (#1) · ⚡ 56.2%
  2. " does". The structure also does not have means for↕ -24.38 (#1) · ⚡ 53.1%
  3. " do" and isolationists who do not want the United↕ -24.25 (#1) · ⚡ 47.5%
  4. " do" in these markets typically do not store information regarding↕ -24.12 (#1) · ⚡ 58.6%
  5. " does" when the operating system does it for you?↕ -23.88 (#1) · ⚡ 49.6%
  6. " do". These designs also do not lend themselves to↕ -23.50 (#1) · ⚡ 53.9%
  7. " does" MySQL cluster. It does read-write splitting↕ -23.50 (#1) · ⚡ 57.8%
  8. " do" Syrian oil fields that do not belong to us↕ -23.38 (#1) · ⚡ 49.2%
  9. " does" etch process generally does not provide precise control↕ -23.25 (#1) · ⚡ 48.6%
  10. " does"_n$ that does not contain a mon↕ -23.12 (#1) · ⚡ 56.6%
  11. " do" the truth. They do not receive↕ -23.12 (#1) · ⚡ 50.4%
  12. " does"savings. It does not disclose using the↕ -23.12 (#1) · ⚡ 50.8%
  13. " does" on one hand it does not cause the temperature↕ -23.00 (#1) · ⚡ 50.4%
  14. " does"This angularJS file does not display the value↕ -23.00 (#1) · ⚡ 52.7%
  15. " do"↵7. They do well in captivity.↕ -23.00 (#1) · ⚡ 53.5%
  16. " do" they are told they do not have enough experience↕ -22.88 (#1) · ⚡ 42.6%
  17. " do" as neutrality or to do anything that might lead↕ -22.88 (#1) · ⚡ 43.9%
  18. " do" 497, we do not recognize Israel�↕ -22.88 (#1) · ⚡ 42.6%
  19. " does"irectional; it does not have a designated↕ -22.75 (#1) · ⚡ 55.5%
  20. " do" the United States to do something about the Israeli↕ -22.75 (#1) · ⚡ 41.4%
GPT-2 Small · L9/C119

Japanese language

Confidence 4/5
Component profile — C119 · layer 9 · positive tail

Logit-lens tokens · positive

#1"�"#2"�"#3"�"#4"し"#5"�"#6"�"#7"�"#8"�"#9"�"#10"�"

R-lens tokens · positive

#1"に"#2"し"#3"�"#4"�"#5"は"#6"�"#7"�"#8"�"#9"�"#10"�"

Skewness

+4.38

Excess kurtosis

49.88rank #110

Logcosh deviation

0.0982rank #120

Top-score occurrences · positive

  1. "�"頭部を強�↕ +18.88 (#1) · ⚡ 42.6%
  2. "�"の午後６�↕ +17.88 (#1) · ⚡ 43.0%
  3. "い"度の高い国会↕ +17.75 (#1) · ⚡ 42.0%
  4. "�"深夜にラーメ↕ +17.62 (#1) · ⚡ 37.5%
  5. "�"が存在した。�↕ +17.38 (#1) · ⚡ 37.9%
  6. "�"の存在につい↕ +17.38 (#1) · ⚡ 35.7%
  7. "�"��構造化↕ +16.88 (#1) · ⚡ 37.3%
  8. "�"�姫一座。もっと↕ +16.88 (#1) · ⚡ 34.8%
  9. "子"る様子をアップ↕ +16.88 (#1) · ⚡ 35.4%
  10. "�"恵夫人とも�↕ +16.88 (#1) · ⚡ 43.4%
  11. "�"知名度を�↕ +16.75 (#1) · ⚡ 41.4%
  12. "�"を共有します↵↕ +16.62 (#1) · ⚡ 22.9%
  13. "�"�時間も見�↕ +16.62 (#1) · ⚡ 32.0%
  14. "�"知名度の高↕ +16.50 (#1) · ⚡ 35.9%
  15. "�"は↵一切なしか…↕ +16.50 (#1) · ⚡ 25.2%
  16. "�"力が前提と↕ +16.50 (#1) · ⚡ 24.6%
  17. "�"推定（�↕ +16.50 (#1) · ⚡ 37.1%
  18. "�"の申年女子た�↕ +16.50 (#1) · ⚡ 27.0%
  19. "�"た数字による↕ +16.50 (#1) · ⚡ 32.6%
  20. "�"現金のみ。↕ +16.50 (#1) · ⚡ 36.3%
GPT-2 Small · L9/C464

Group/list

Confidence 5/5
Component profile — C464 · layer 9 · negative tail

Logit-lens tokens · negative

#1" consisting"#2" comprising"#3" comprised"#4" members"#5" member"#6"members"#7"ings"#8" composed"#9"mates"#10" mates"

R-lens tokens · negative

#1" consisting"#2" members"#3" member"#4" composed"#5" comprising"#6" comprised"#7"members"#8" membership"#9" composition"#10"rons"

Skewness

-1.56

Excess kurtosis

11.03rank #525

Logcosh deviation

0.0510rank #465

Top-score occurrences · negative

  1. "Group"-2025 ("Group I plaintiffs") filed↕ -13.69 (#1) · ⚡ 20.1%
  2. " group"on-one or group play, encouraging players↕ -13.50 (#1) · ⚡ 14.7%
  3. " Group" than one person. Group (or co-)↕ -13.25 (#1) · ⚡ 21.1%
  4. "List" adapter.setList(eventsResults);↕ -13.19 (#1) · ⚡ 23.5%
  5. " group" miss occurs, a group of pieces of data↕ -13.19 (#1) · ⚡ 26.2%
  6. " list":↵↵Getting list of Variables of↕ -13.00 (#1) · ⚡ 23.3%
  7. " group" the users or a group of users or relay↕ -12.94 (#1) · ⚡ 22.9%
  8. " List". I have a ListView with Label and↕ -12.94 (#1) · ⚡ 21.0%
  9. " List"Python For Loop using List - Trying to group↕ -12.88 (#1) · ⚡ 21.7%
  10. "Group"A:↵↵Group by the second and↕ -12.81 (#1) · ⚡ 19.5%
  11. " list" button event with a list view not getting the↕ -12.75 (#1) · ⚡ 23.9%
  12. " List" public List<ScaleTransactionDet↕ -12.69 (#1) · ⚡ 20.3%
  13. "list"SuperTypeList(list: KtSuper↕ -12.69 (#1) · ⚡ 19.7%
  14. " List" public List<T> find↕ -12.62 (#1) · ⚡ 16.3%
  15. " list"_data in a list↵pd.Data↕ -12.62 (#1) · ⚡ 24.2%
  16. " group"uses are a large group of positive-str↕ -12.62 (#1) · ⚡ 20.7%
  17. " group" which is gang or group, rather than calling↕ -12.56 (#1) · ⚡ 18.4%
  18. " group"↵// of a group.↵message AP↕ -12.56 (#1) · ⚡ 21.5%
  19. " group" row number differently for group try =RowNumber↕ -12.44 (#1) · ⚡ 21.4%
  20. "list"↵#include <list>↵#include↕ -12.44 (#1) · ⚡ 17.9%
GPT-2 Small · L9/C619

Way/approach descriptor

Confidence 4/5
Component profile — C619 · layer 9 · positive tail

Logit-lens tokens · positive

#1" way"#2" solution"#3" alternative"#4" method"#5" workaround"#6" solutions"#7" ways"#8" alternatives"#9" remedy"#10" approach"

R-lens tokens · positive

#1" solution"#2" way"#3" alternative"#4" method"#5" solutions"#6" workaround"#7" alternatives"#8" ways"#9" option"#10" methods"

Skewness

+1.00

Excess kurtosis

7.15rank #612

Logcosh deviation

0.0347rank #620

Top-score occurrences · positive

  1. " easiest" block.↵The easiest way is to send↕ +16.38 (#1) · ⚡ 23.4%
  2. " only" assembly yourself is the only real approach that is↕ +15.94 (#1) · ⚡ 23.5%
  3. " alternative", is there any alternative to achieve the same↕ +15.44 (#1) · ⚡ 23.8%
  4. " viable"wing may be a viable alternative to brewing.↕ +15.12 (#1) · ⚡ 23.7%
  5. " better"↵Is there any better way? I already↕ +14.81 (#1) · ⚡ 22.2%
  6. " efficient" What is the most efficient way?↵I↕ +14.75 (#1) · ⚡ 22.0%
  7. " best" queries. The best you can do is↕ +14.75 (#1) · ⚡ 18.0%
  8. " better" Or is there any better practice to keep this↕ +14.44 (#1) · ⚡ 20.8%
  9. " simplest" database engines. The simplest solution for your task↕ +14.31 (#1) · ⚡ 20.4%
  10. " efficient" see that the most efficient way to redirect and↕ +14.31 (#1) · ⚡ 21.2%
  11. " effective"s could be an effective preventive strategy against various↕ +14.25 (#1) · ⚡ 21.8%
  12. " safest" seems to be the safest and most cost-↕ +14.12 (#1) · ⚡ 19.9%
  13. " easiest" stream.↵The easiest solution is to wrap↕ +14.00 (#1) · ⚡ 19.1%
  14. " best" so what is the best approach: use another↕ +13.88 (#2) · ⚡ 19.3%
  15. " better" maybe there's a better way to do this↕ +13.75 (#1) · ⚡ 20.2%
  16. " potential" demonstrates promise as a potential intervention in this domain↕ +13.56 (#1) · ⚡ 21.0%
  17. " simplest" yearly etc. The simplest thing I have done↕ +13.50 (#1) · ⚡ 18.4%
  18. " only" Davies knew that the only line of withdrawal lay↕ +13.38 (#1) · ⚡ 14.9%
  19. " alternate" could think of an alternate @param tag like↕ +13.31 (#1) · ⚡ 19.6%
  20. " efficient"ers, is an efficient procedure with associated high↕ +13.12 (#1) · ⚡ 19.0%
GPT-2 Small · L9/C730

Who

Confidence 5/5
Component profile — C730 · layer 9 · positive tail

Logit-lens tokens · positive

#1" pissed"#2"anwhile"#3" vacated"#4"upon"#5" stepped"#6" owns"#7" replaced"#8" joined"#9" boarded"#10" meanwhile"

R-lens tokens · positive

#1" pissed"#2" rightfully"#3" vacated"#4"anwhile"#5" owns"#6" stepped"#7" superv"#8" meanwhile"#9" promptly"#10" downed"

Skewness

+0.25

Excess kurtosis

2.21rank #731

Logcosh deviation

0.0164rank #731

Top-score occurrences · positive

  1. "who" friend Sumika (who also got admitted to↕ +12.56 (#1) · ⚡ 17.3%
  2. " who" ex-husband, who lives here and owns↕ +11.38 (#1) · ⚡ 16.9%
  3. " who" Stephen Strulens who are of the opposite↕ +11.00 (#1) · ⚡ 16.0%
  4. " who"inda Rushon, who was the subpoena clerk↕ +10.94 (#1) · ⚡ 15.5%
  5. " who" Seydel, who are involved with a↕ +10.81 (#1) · ⚡ 14.2%
  6. " who"year-old companion who had stepped out of↕ +10.69 (#2) · ⚡ 12.7%
  7. " who" White, Jr., who practices in London,↕ +10.62 (#1) · ⚡ 15.0%
  8. " who" official Wendy Nichols, who was chairing the↕ +10.25 (#1) · ⚡ 12.2%
  9. " who"ashvili, who earlier had a mid↕ +10.25 (#1) · ⚡ 16.0%
  10. " who", and Berman, who announced plans last month↕ +10.19 (#2) · ⚡ 10.1%
  11. " who"man Sporn, who advises Mr. Cruz↕ +10.06 (#1) · ⚡ 12.2%
  12. " who"chagrin, who lives near the bar↕ +9.81 (#1) · ⚡ 11.7%
  13. " who"D-MA), who remain neck and neck↕ +9.56 (#2) · ⚡ 8.7%
  14. " who" chief of staff, who in the post-↕ +9.50 (#1) · ⚡ 12.3%
  15. " who" Eitel, who then killed himself in↕ +9.50 (#2) · ⚡ 11.3%
  16. " who"�Rourke, who wouldn’t↕ +9.38 (#1) · ⚡ 11.2%
  17. " who" drummer Gavin Harrison, who is highly regarded as↕ +9.31 (#1) · ⚡ 12.8%
  18. "who" with McWilliams (who had illustrated some of↕ +9.25 (#1) · ⚡ 11.8%
  19. " who"en Dunst, who co-starred↕ +9.25 (#1) · ⚡ 11.3%
  20. " who"aging comments from people who should know better,↕ +9.19 (#1) · ⚡ 13.2%
GPT-2 Small · L10/C294

Suffix comp-

Confidence 5/5
Component profile — C294 · layer 10 · negative tail

Logit-lens tokens · negative

#1"ression"#2"onents"#3"artment"#4"ressor"#5"icit"#6"ulsion"#7"ension"#8"ention"#9"resso"#10"lections"

R-lens tokens · negative

#1"ression"#2"resso"#3"ressor"#4"onents"#5"ention"#6"lections"#7"ension"#8"endant"#9"ressed"#10"artment"

Skewness

-2.21

Excess kurtosis

17.61rank #372

Logcosh deviation

0.0659rank #295

Top-score occurrences · negative

  1. " Comp"): T; // Compiler error↵end↕ -14.31 (#1) · ⚡ 17.0%
  2. "Comp"latorImpl.doCompile(QueryTrans↕ -14.19 (#1) · ⚡ 17.7%
  3. "Comp"↵• “Competition” Mode↕ -14.12 (#1) · ⚡ 16.6%
  4. "Comp".↵↵"Competition for AI superiority↕ -13.75 (#1) · ⚡ 14.7%
  5. "Comp" and cities↵↵Compiling campaign metrics for↕ -13.69 (#1) · ⚡ 15.5%
  6. "Comp" Application.SetCompatibleTextRend↕ -13.69 (#1) · ⚡ 17.6%
  7. "Comp"ribeDrdsComponentsRequest(R↕ -13.62 (#1) · ⚡ 19.3%
  8. "comp" but as a "compiled" data store↕ -13.56 (#1) · ⚡ 16.2%
  9. "comp"disposing && (components != null))↕ -13.56 (#1) · ⚡ 17.4%
  10. "Comp"for example, "Compulsory Heter↕ -13.56 (#1) · ⚡ 17.0%
  11. "comp"ers_and_components/creating↕ -13.50 (#1) · ⚡ 17.0%
  12. "Comp" m_RemovedComponents: []↵↕ -13.44 (#1) · ⚡ 17.6%
  13. "Def" States↵Category:Defunct companies based in↕ -13.44 (#1) · ⚡ 15.6%
  14. "Rep"entially.↵↵Repetition and contrast are↕ -13.38 (#1) · ⚡ 17.8%
  15. "Supp")?↵458↵Suppose -s =↕ -13.38 (#1) · ⚡ 19.7%
  16. "Def" Appellant (Defendant),↵v↕ -13.38 (#1) · ⚡ 14.1%
  17. "Comp" regexp.MustCompile("\\n↕ -13.31 (#1) · ⚡ 15.4%
  18. "comp"namespace name="components" file="/↕ -13.25 (#1) · ⚡ 19.0%
  19. "Def" Boston↵Category:Defunct manufacturing companies of↕ -13.25 (#1) · ⚡ 15.8%
  20. "Ext" right.↵↵Extensive interviews with several↕ -13.25 (#1) · ⚡ 14.0%
GPT-2 Small · L10/C307

Opening parentheses in code

Confidence 5/5
Component profile — C307 · layer 10 · positive tail

Logit-lens tokens · positive

#1"new"#2"true"#3"input"#4"const"#5"this"#6"null"#7"std"#8"false"#9"data"#10"args"

R-lens tokens · positive

#1"escription"#2"args"#3"new"#4"target"#5"�"#6"uid"#7"data"#8"input"#9"array"#10"�"

Skewness

+2.43

Excess kurtosis

20.42rank #306

Logcosh deviation

0.0647rank #308

Top-score occurrences · positive

  1. "("_from_word(word)↵ ↕ +13.50 (#1) · ⚡ 23.1%
  2. "("tSolverName(arg);↵ ↕ +13.25 (#1) · ⚡ 20.8%
  3. "(" new CStringEntry(0, (const↕ +13.19 (#1) · ⚡ 23.9%
  4. "("animal = new T(5);↵ ↕ +13.00 (#1) · ⚡ 23.6%
  5. "("inet_ntoa(struct.pack('↕ +12.81 (#1) · ⚡ 21.1%
  6. "("UserByUsername(username,function(↕ +12.75 (#1) · ⚡ 21.8%
  7. "(" friend int ::main(int argc,↕ +12.69 (#1) · ⚡ 19.6%
  8. "("apping.fromJSON(json, {↵↕ +12.69 (#1) · ⚡ 21.5%
  9. "("VarSubstCopy(SgVariableSy↕ +12.62 (#1) · ⚡ 21.0%
  10. "("SwapOnRead(val.at("↕ +12.62 (#1) · ⚡ 22.2%
  11. "(" base.SelectTemplate(item, container);↕ +12.56 (#1) · ⚡ 24.7%
  12. "(" JSON.stringify(newOptions.body↕ +12.56 (#1) · ⚡ 21.0%
  13. "("Instance.ToJSON(value, _param↕ +12.44 (#1) · ⚡ 19.4%
  14. "("->readDatagram(datagram.data↕ +12.44 (#1) · ⚡ 21.7%
  15. "("logger.Print(v...)↵}↕ +12.44 (#1) · ⚡ 15.7%
  16. "(" intent = new Intent(this, MyClass↕ +12.44 (#1) · ⚡ 20.2%
  17. "(".getTeamEvents(entity.getTeam↕ +12.38 (#1) · ⚡ 18.3%
  18. "("OptionsItemSelected(item);↵}↕ +12.31 (#1) · ⚡ 19.3%
  19. "("Instance.ToObject(new StreamReader(↕ +12.25 (#1) · ⚡ 21.7%
  20. "("ression = new Expression(Class.class,↕ +12.25 (#1) · ⚡ 20.4%
GPT-2 Small · L10/C445

Expecting a temporal-duration phrase

Confidence 4/5
Component profile — C445 · layer 10 · negative tail

Logit-lens tokens · negative

#1" since"#2" lately"#3" throughout"#4"since"#5" recently"#6" extensively"#7" over"#8" steadily"#9" ever"#10" intermitt"

R-lens tokens · negative

#1" since"#2" lately"#3" throughout"#4" recently"#5" extensively"#6"since"#7" over"#8" intermitt"#9" steadily"#10" previously"

Skewness

-1.67

Excess kurtosis

13.63rank #447

Logcosh deviation

0.0495rank #446

Top-score occurrences · negative

  1. " firm" has been with the firm since 2001. Both↕ -21.25 (#1) · ⚡ 27.5%
  2. " on" stuff has been going on for↵well over↕ -19.50 (#1) · ⚡ 26.8%
  3. " making" that he has been making since the early days↕ -18.38 (#1) · ⚡ 23.7%
  4. " Congress". Republicans have controlled Congress completely for 10 of↕ -18.25 (#1) · ⚡ 23.8%
  5. " saying"’ve been saying for years. The↕ -17.75 (#1) · ⚡ 27.1%
  6. " diabetic" "I've been diabetic for 20 years and↕ -17.75 (#1) · ⚡ 23.6%
  7. "and" in custody on remand since mid-April↕ -17.38 (#1) · ⚡ 22.1%
  8. " us" crew have been with us in other similar↵↕ -17.38 (#1) · ⚡ 32.4%
  9. " works" had been in the works for “months↕ -17.00 (#1) · ⚡ 22.1%
  10. " government" control of the Syrian government since seizing it from↕ -17.00 (#1) · ⚡ 26.2%
  11. " God" have been praying to God to ask for help↕ -16.62 (#1) · ⚡ 21.0%
  12. " around", which has been around for two decades,↕ -16.62 (#1) · ⚡ 23.4%
  13. "RS" has worked at SRS more than 30 years↕ -16.50 (#1) · ⚡ 21.4%
  14. " other" wife had known each other about 18 months when↕ -16.50 (#1) · ⚡ 19.6%
  15. " customers" has been available to customers for more than three↕ -16.50 (#1) · ⚡ 25.2%
  16. " beings"ths have plagued human beings even before recorded history↕ -16.12 (#1) · ⚡ 26.2%
  17. " us" venues that have supported us in the past.↕ -16.12 (#1) · ⚡ 27.7%
  18. " there"’ve lived there for a decade,↕ -16.00 (#1) · ⚡ 22.3%
  19. " ever" the doldrums ever since.↵↵↕ -15.75 (#1) · ⚡ 18.5%
  20. " region" has been in the region for quite some time↕ -15.56 (#1) · ⚡ 24.4%
GPT-2 Small · L10/C446

Opening quotation mark

Confidence 5/5
Component profile — C446 · layer 10 · positive tail

Logit-lens tokens · positive

#1"nothing"#2"every"#3"everything"#4"what"#5"the"#6"little"#7"many"#8"they"#9"there"#10"their"

R-lens tokens · positive

#1"embedreportprint"#2" externalToEVA"#3"\u000f"#4"\u001a"#5"�"#6"�"#7"\u0003"#8"\t"#9" RandomRedditor"#10"\u0001"

Skewness

+1.79

Excess kurtosis

15.26rank #416

Logcosh deviation

0.0494rank #447

Top-score occurrences · positive

  1. " \""isprudence has "grown like a house↕ +15.75 (#1) · ⚡ 23.6%
  2. " '" she pushed back she 'took her fist↕ +15.38 (#1) · ⚡ 18.4%
  3. " \"" the remaining personnel could "enjoy peace of↕ +15.19 (#1) · ⚡ 22.2%
  4. " \""en said she had "an enormous reservoir of↕ +15.12 (#1) · ⚡ 22.8%
  5. " \"", said the club "can really achieve something↕ +15.12 (#1) · ⚡ 20.3%
  6. " \"" herself said she had "every anticipation" she↕ +15.00 (#1) · ⚡ 21.8%
  7. " \"" Steven Mnuchin, "also has promised to↕ +15.00 (#1) · ⚡ 21.2%
  8. " \"" Times, that he "felt it was important↕ +14.75 (#1) · ⚡ 25.2%
  9. " \"" out that it is "an important foundation industry↕ +14.69 (#1) · ⚡ 22.6%
  10. " \"" Russia on cybersecurity to "partnering with↕ +14.62 (#1) · ⚡ 18.6%
  11. " \"" accused Fine Gael of "pulling rubber levers↕ +14.62 (#1) · ⚡ 21.1%
  12. " \"" Brothers - who were "on↵a mission↕ +14.62 (#1) · ⚡ 26.0%
  13. " \"" biodiversity there is still "mostly unexplored and↕ +14.62 (#1) · ⚡ 23.1%
  14. " \"" among other things, "stood with Republican extremists↕ +14.62 (#1) · ⚡ 19.8%
  15. " \"" Iris Miller says they "lived better than the↕ +14.56 (#1) · ⚡ 20.9%
  16. " \"" the new system could "potentially" lead↕ +14.50 (#1) · ⚡ 21.7%
  17. " \""grown" and had "not been touched in↕ +14.44 (#1) · ⚡ 20.6%
  18. " \"" said the growth figures "could have been worse↕ +14.44 (#1) · ⚡ 17.5%
  19. " \"" Commons the decisions were "painful" but↕ +14.44 (#1) · ⚡ 21.0%
  20. " \"" demanded that Brown get "full justice, not↕ +14.38 (#1) · ⚡ 18.3%
GPT-2 Small · L10/C493

Occupational title or official role

Confidence 5/5
Component profile — C493 · layer 10 · negative tail

Logit-lens tokens · negative

#1" extraord"#2"ial"#3" specializing"#4"esses"#5" emer"#6"ials"#7"ially"#8"hips"#9" overseeing"#10"dom"

R-lens tokens · negative

#1" extraord"#2"ial"#3"hips"#4"ials"#5" specializing"#6"ially"#7" emer"#8" overseeing"#9"esses"#10" who"

Skewness

-1.42

Excess kurtosis

10.41rank #518

Logcosh deviation

0.0442rank #494

Top-score occurrences · negative

  1. " Constable" 2009, the Chief Constable of the Police Service↕ -13.06 (#1) · ⚡ 17.5%
  2. " Governor"1951), acting Governor of New Jersey in↕ -12.75 (#1) · ⚡ 20.1%
  3. " Speaker" 1966) was the Speaker of Parliament in Kyr↕ -12.56 (#1) · ⚡ 19.6%
  4. " editor"er, the medical editor for ABC News,↕ -12.38 (#1) · ⚡ 19.8%
  5. "President"↵↵Vice-President↵↵1308↕ -12.25 (#1) · ⚡ 19.2%
  6. " Editor" of the first City Editor of the Los Angeles↕ -12.19 (#1) · ⚡ 16.2%
  7. " officer" was the chief HR officer for SRI International↕ -12.06 (#1) · ⚡ 21.2%
  8. " Director" assumed the Chief Executive Director (CED)↕ -12.00 (#1) · ⚡ 11.2%
  9. " secretary" legal counsel and corporate secretary at Richemont↕ -12.00 (#1) · ⚡ 17.1%
  10. " professor"orman, a communications professor at MacEwan↕ -11.94 (#1) · ⚡ 14.9%
  11. " editor" The other was fiction editor Vicky Chen Ha↕ -11.88 (#2) · ⚡ 14.1%
  12. " coach" an assistant defensive backs coach in 2016, but↕ -11.88 (#1) · ⚡ 13.4%
  13. " Attorney" Pines, City Attorney, William B.↕ -11.88 (#1) · ⚡ 22.3%
  14. " Attorney"↵Assistant United States Attorney, Charlotte, North↕ -11.88 (#1) · ⚡ 16.8%
  15. " Designer" Design and Deputy Head Designer at Bugatti,↕ -11.88 (#1) · ⚡ 13.9%
  16. " President" as Microsoft's Vice President of Federal Government business↕ -11.81 (#1) · ⚡ 14.2%
  17. " editor" Assistant Editor, becoming editor-in-Chief↕ -11.81 (#1) · ⚡ 13.8%
  18. "or", then-governor of Massachusetts and chairman↕ -11.81 (#1) · ⚡ 15.9%
  19. "or".↵↵Governor-General↵The↕ -11.75 (#1) · ⚡ 15.6%
  20. " governor" know, is now governor of the Bank of↕ -11.69 (#1) · ⚡ 19.9%
GPT-2 Small · L11/C243

Expecting letter D but need more conditions

Confidence 3/5
Component profile — C243 · layer 11 · positive tail

Logit-lens tokens · positive

#1"D"#2"d"#3"dt"#4"ds"#5"dp"#6"DD"#7" D"#8"Ds"#9"DP"#10"DL"

R-lens tokens · positive

#1"D"#2"d"#3"dt"#4"ds"#5"dp"#6"DD"#7" D"#8"Ds"#9"DP"#10"DL"

Skewness

+2.12

Excess kurtosis

18.73rank #277

Logcosh deviation

0.0636rank #244

Top-score occurrences · positive

  1. "1"-name="step1d" title="↕ +23.00 (#1) · ⚡ 40.8%
  2. "_"↵#define SQL_DATATY↕ +22.75 (#1) · ⚡ 34.8%
  3. "_"↵#define SQL_DATATY↕ +20.75 (#1) · ⚡ 34.8%
  4. " 2"" title="Step 2d" style="↕ +20.25 (#1) · ⚡ 35.5%
  5. " us" be purchased for 5 usd He↕ +18.75 (#1) · ⚡ 37.7%
  6. "HR" Resources Development (IHRD) is an↕ +18.12 (#1) · ⚡ 31.2%
  7. "o"bRi_oDDAc}↕ +17.25 (#1) · ⚡ 29.7%
  8. "Up" ' + fechaUpd + ' -↕ +17.25 (#1) · ⚡ 32.4%
  9. ".", not the O.D. of the↕ +17.25 (#1) · ⚡ 30.5%
  10. "Up" $.PhotoUpdater.do↕ +17.12 (#1) · ⚡ 32.4%
  11. "{"- \dot{d} _↕ +17.12 (#1) · ⚡ 35.5%
  12. "*"?↵-395*d**3 +↕ +17.00 (#1) · ⚡ 37.1%
  13. "00" "\x00DZD\↕ +17.00 (#1) · ⚡ 25.6%
  14. " us" be purchased for 5 usd He won't↕ +17.00 (#1) · ⚡ 32.4%
  15. "409"5ff0809409d15d8↕ +16.88 (#1) · ⚡ 32.6%
  16. "45" 00140 808945E0 3C↕ +16.75 (#1) · ⚡ 18.9%
  17. "ns"↵ Waßmannsdorf Fire Service↕ +16.38 (#1) · ⚡ 33.6%
  18. "*"**4 - 3*d**2 -↕ +16.25 (#1) · ⚡ 36.9%
  19. " GET"([DOD], GETDATE()) /↕ +16.25 (#1) · ⚡ 28.1%
  20. ".("dA)50.(dT)50↕ +16.12 (#1) · ⚡ 28.7%
GPT-2 Small · L11/C593

Official agency or department

Confidence 4/5
Component profile — C593 · layer 11 · positive tail

Logit-lens tokens · positive

#1" office"#2" offices"#3" headquarters"#4" official"#5" Office"#6" adviser"#7" agency"#8" officials"#9" officer"#10" inspector"

R-lens tokens · positive

#1" office"#2" offices"#3" headquarters"#4" official"#5" Office"#6" adviser"#7" agency"#8" officials"#9" officer"#10" inspector"

Skewness

+0.83

Excess kurtosis

5.68rank #610

Logcosh deviation

0.0313rank #594

Top-score occurrences · positive

  1. " Security" Public Safety and Homeland Security Bureau from November 2013↕ +15.19 (#1) · ⚡ 18.6%
  2. " defence"ino," Colombia's defence ministry said in a↕ +15.06 (#1) · ⚡ 14.4%
  3. " Defense" have a Department of Defense,” Gork↕ +14.94 (#1) · ⚡ 17.3%
  4. " defence"↵↵Russia's defence ministry denied it was↕ +14.75 (#1) · ⚡ 14.7%
  5. " Commerce", a Ministry of Commerce spokesman declined to explain↕ +14.38 (#1) · ⚡ 15.6%
  6. " treasury"But as former assistant treasury secretary for financial institutions↕ +14.31 (#1) · ⚡ 15.3%
  7. " ministry"ao Feng, a ministry spokesman, told reporters↕ +14.06 (#1) · ⚡ 17.0%
  8. " police"rick Train, a police regional director, said↕ +13.31 (#1) · ⚡ 13.9%
  9. " defense" the role of acting defense secretary since Mattis'↕ +13.12 (#1) · ⚡ 14.0%
  10. " finance" under development , a finance ministry statement said.↕ +13.00 (#1) · ⚡ 12.2%
  11. "intelligence"’s counterintelligence analysts about precisely how↕ +12.75 (#1) · ⚡ 17.3%
  12. " Operations"als Intelligence and Network Operations Group, and a↕ +12.50 (#1) · ⚡ 12.6%
  13. " security"’s former security advisers deemed the drone↕ +12.38 (#1) · ⚡ 14.6%
  14. " security" Monday, the national security council’s↕ +12.19 (#1) · ⚡ 14.4%
  15. " police"ma: A special police officer, he conducted↕ +12.12 (#1) · ⚡ 12.1%
  16. " Defense"!"↵↵German Defense Minister Ursula von↕ +12.06 (#1) · ⚡ 11.9%
  17. " legal"y Henderson, DNC legal consultant.↵↵↕ +12.06 (#1) · ⚡ 12.1%
  18. " military" deaths. A UK military spokesman said: "↕ +12.00 (#1) · ⚡ 10.7%
  19. " House" a former senior White House official told Peter Baker↕ +11.81 (#1) · ⚡ 9.9%
  20. " DHS"More than million in DHS contracts awarded to firm↕ +11.81 (#1) · ⚡ 11.1%
GPT-2 Small · L11/C752

?

Confidence 0/5
Component profile — C752 · layer 11 · negative tail

Logit-lens tokens · negative

#1"opped"#2"aji"#3"ohan"#4"NetMessage"#5"�"#6"quin"#7"�"#8"cloneembedreportprint"#9"wall"#10"our"

R-lens tokens · negative

#1"opped"#2"aji"#3"ohan"#4"NetMessage"#5"�"#6"quin"#7"�"#8"cloneembedreportprint"#9"wall"#10"our"

Skewness

-0.02

Excess kurtosis

0.62rank #755

Logcosh deviation

0.0089rank #753

Top-score occurrences · negative

  1. "-" hilarious coming-of-age comedy that stars↕ -6.66 (#2) · ⚡ 2.8%
  2. " hung" families as President Obama hung the blue ribbon cr↕ -6.28 (#1) · ⚡ 4.4%
  3. "-" another coming-of-age film of sorts↕ -6.16 (#3) · ⚡ 2.7%
  4. " into" time of the coming into force of this Act↕ -6.03 (#3) · ⚡ 1.9%
  5. "fl" when he flip-flops on crucial issues↕ -6.00 (#2) · ⚡ 2.9%
  6. " bl" occurred. The embarrassing blunder involves Department of↕ -5.97 (#2) · ⚡ 2.4%
  7. " was" after the Paris Agreement was adopted, the six↕ -5.91 (#4) · ⚡ 2.1%
  8. "-" Political Participation in Post-Communist Europe.↕ -5.88 (#5) · ⚡ 2.3%
  9. " pl" swim against orthodoxy by plumping for a draw↕ -5.84 (#2) · ⚡ 2.6%
  10. " of"�as an act of exploitation,”↕ -5.62 (#3) · ⚡ 2.4%
  11. " of" mercy and a sign of respect (cf.↕ -5.59 (#4) · ⚡ 2.4%
  12. "fl" judgment about flip-flopping; we're↕ -5.59 (#2) · ⚡ 3.1%
  13. " in" upon the entire record in a prior appeal.↕ -5.56 (#4) · ⚡ 3.1%
  14. " African" set a new South African 400m freestyle↕ -5.47 (#6) · ⚡ 1.4%
  15. " actual" do not reflect my actual views and have no↕ -5.47 (#3) · ⚡ 2.3%
  16. " Dynamo"old is the second Dynamo player within a couple↕ -5.44 (#8) · ⚡ 1.5%
  17. " post" an intraoperative and postoperative continued IVLI↕ -5.38 (#5) · ⚡ 2.5%
  18. " may" this topic? You may also like these photo↕ -5.38 (#2) · ⚡ 2.2%
  19. " 1956" born March 2, 1956 in Vrbas↕ -5.31 (#3) · ⚡ 2.9%
  20. " in" wrote the opinion, in which Judge WILLI↕ -5.31 (#3) · ⚡ 2.6%
Gemma 2 2B · L0/C603

Pressure

Confidence 5/5
Component profile — C603 · layer 0 · positive tail

Logit-lens tokens · positive

#1" pressure"#2" Pressure"#3"pressure"#4"Pressure"#5" PRESSURE"#6" pressures"#7" pressured"#8" pression"#9" presión"#10" pressão"

R-lens tokens · positive

#1" pressure"#2" Pressure"#3" PRESSURE"#4"Pressure"#5"pressure"#6" pressures"#7" pression"#8" pressured"#9" pressão"#10" presión"

Skewness

+24.53

Excess kurtosis

1317.27rank #183

Logcosh deviation

0.1860rank #604

Top-score occurrences · positive

  1. " pressure" when electrokinetic and pressure-driven particle velocities↕ +59.75 (#1) · ⚡ 77.3%
  2. " pressure" to as strain and pressure sensors, can collect↕ +59.25 (#1) · ⚡ 81.6%
  3. " pressure" with conversion and total pressure. As the water↕ +58.50 (#1) · ⚡ 81.2%
  4. " pressure" the groups). The pressure-volume diagram showed↕ +58.50 (#1) · ⚡ 79.7%
  5. " pressure" is defined between the pressure tube and a reservoir↕ +58.50 (#1) · ⚡ 80.5%
  6. " pressure" and employing the resultant pressure wave for producing fine↕ +58.50 (#1) · ⚡ 82.4%
  7. " pressure" a common manifold absolute pressure (MAP) of↕ +58.25 (#1) · ⚡ 81.6%
  8. " pressure" out under a reduced pressure atmosphere evacuated by a↕ +58.25 (#1) · ⚡ 79.7%
  9. " pressure" subsystem to process the pressure signals, a counter↕ +58.25 (#1) · ⚡ 82.0%
  10. " pressure" fluid. When the pressure is transmitted to the↕ +58.25 (#1) · ⚡ 80.5%
  11. " pressure" trans-stenotic pressure gradient of over ↕ +58.25 (#1) · ⚡ 81.6%
  12. " pressure" In contrast, pulse pressure variation, variation of↕ +58.25 (#1) · ⚡ 81.6%
  13. " pressure" raising the metal vapor pressure above 5 atmospheres↕ +58.25 (#1) · ⚡ 78.1%
  14. " pressure" (and corresponding clamping pressure) concurrent with reducing↕ +58.25 (#1) · ⚡ 80.5%
  15. " pressure"Effect of elevated hydrostatic pressure on the shape of↕ +58.25 (#1) · ⚡ 78.5%
  16. " pressure" clearly generates a significant pressure on the FIFO circuits↕ +58.25 (#1) · ⚡ 82.0%
  17. " pressure" obtain a large sound pressure. As a result↕ +58.25 (#1) · ⚡ 78.1%
  18. " pressure" and downward by the pressure of the expanded working↕ +58.25 (#1) · ⚡ 80.9%
  19. " pressure"/// derived dimension for pressure : L^-1↕ +58.00 (#1) · ⚡ 75.4%
  20. " pressure" system which incorporates a pressure balancer and a high↕ +58.00 (#1) · ⚡ 82.8%
Gemma 2 2B · L0/C761

3rd-person singular present verb

Confidence 5/5
Component profile — C761 · layer 0 · negative tail

Logit-lens tokens · negative

#1" itſelf"#2" myſelf"#3" Theſe"#4" Monfieur"#5" Anſ"#6" ―――――"#7" iſt"#8" doubtnut"#9"ſelf"#10" raiſ"

R-lens tokens · negative

#1"Adds"#2" adds"#3" builds"#4" gets"#5" learns"#6" engages"#7" theres"#8"Opens"#9" shes"#10" delivers"

Skewness

-6.38

Excess kurtosis

68.75rank #1569

Logcosh deviation

0.1688rank #762

Top-score occurrences · negative

  1. " finds" temper. Kirishima finds out the reason why↕ -15.75 (#2) · ⚡ 9.2%
  2. " compares" plate type of binding compares favorable with that of↕ -15.62 (#2) · ⚡ 5.1%
  3. " goes" involved.”↵↵Ford goes on to say that↕ -15.62 (#2) · ⚡ 9.1%
  4. " takes".Valerie Solanas takes a .32↕ -15.44 (#2) · ⚡ 10.9%
  5. " takes" is over and Adama takes a moment to tell↕ -15.44 (#2) · ⚡ 10.8%
  6. " adds"99," Sullivan adds, and "no↕ -15.44 (#3) · ⚡ 6.7%
  7. " adds" not fully understood,” adds Prof. Dean,↕ -15.44 (#3) · ⚡ 6.6%
  8. " puts" shin pain. That puts me at:↵↵↕ -15.38 (#2) · ⚡ 7.2%
  9. " takes", Heather Mac Donald takes on identity politics far↕ -15.38 (#2) · ⚡ 9.7%
  10. " puts" reversing Conservative legislation that puts the final call in↕ -15.31 (#2) · ⚡ 7.0%
  11. " finds" his grandfather, Daphne finds herself at a crossroads↕ -15.25 (#2) · ⚡ 8.4%
  12. " finds" powersuit, she finds herself cast down to↕ -15.25 (#2) · ⚡ 8.9%
  13. " goes" pimp. Fiddler goes after him, and↕ -15.12 (#2) · ⚡ 9.1%
  14. " creates" U.S. creates large fiscal burdens for↕ -15.06 (#2) · ⚡ 10.4%
  15. " goes" shenanigans, Duiker goes to a meeting with↕ -15.06 (#2) · ⚡ 9.1%
  16. " goes" market-basket currency goes back to at least↕ -15.06 (#2) · ⚡ 9.3%
  17. " goes".↵↵The report goes on to assert that↕ -15.06 (#2) · ⚡ 8.9%
  18. " gets"atevs. Tami gets annoyed in the car↕ -15.00 (#2) · ⚡ 12.4%
  19. " goes" But before Gran Torino goes full instructor on him↕ -15.00 (#2) · ⚡ 8.9%
  20. " gets" fiction.↵↵Wood gets into all those issues↕ -14.94 (#2) · ⚡ 12.9%
Gemma 2 2B · L0/C1423

Australian

Confidence 5/5
Component profile — C1423 · layer 0 · negative tail

Logit-lens tokens · negative

#1" Australian"#2"Australian"#3" Australia"#4"Australia"#5" Australians"#6" Aussie"#7" Australien"#8" australian"#9" australiano"#10" AUSTRALIA"

R-lens tokens · negative

#1" Australian"#2"Australian"#3" Australians"#4" Australia"#5" australiano"#6"Australia"#7" AUSTRALIA"#8" australian"#9" Australien"#10" AUSTR"

Skewness

-3.14

Excess kurtosis

109.27rank #1406

Logcosh deviation

0.1086rank #1424

Top-score occurrences · negative

  1. " Australian" Members of the Western Australian Legislative Council, ↕ -36.75 (#1) · ⚡ 27.7%
  2. " Australian"Billionaire returns top Australian award↵↵February ↕ -35.25 (#1) · ⚡ 31.4%
  3. " Australian" replace the South East Australian Basketball League (SE↕ -35.00 (#1) · ⚡ 27.7%
  4. " Australian" time to peg the Australian dollar?↵↵The↕ -35.00 (#1) · ⚡ 32.2%
  5. " Australian"- transparent color, Australian Made triangular corflute↕ -35.00 (#1) · ⚡ 31.6%
  6. " Australian" of great books by Australian Sci-Fi,↕ -35.00 (#1) · ⚡ 32.2%
  7. " Australian" government."↵↵The Australian Communications and Media Authority↕ -34.75 (#1) · ⚡ 31.2%
  8. " Australian" chief executive of the Australian Council for International Development↕ -34.75 (#1) · ⚡ 31.4%
  9. " Australian" largest such payout in Australian legal history.↵↵↕ -34.75 (#1) · ⚡ 32.6%
  10. " Australian" were assessed using the Australian Eating Survey (AES↕ -34.75 (#1) · ⚡ 30.7%
  11. " Australian"-year-old Australian Shepherd who lives with↕ -34.75 (#1) · ⚡ 32.0%
  12. " Australian"8 bottle of red Australian wine and a filet↕ -34.75 (#1) · ⚡ 32.2%
  13. " Australian" in Newman in the Australian state of Western Australia↕ -34.75 (#1) · ⚡ 32.0%
  14. " Australian"10th Annual Australian Club Entertainment Awards↵↵↕ -34.75 (#1) · ⚡ 32.6%
  15. " Australian"0am↵↵An Australian billionaire philanthropist returned the↕ -34.50 (#1) · ⚡ 31.6%
  16. " Australian" owned by the North Australian Pastoral Company since ↕ -34.50 (#1) · ⚡ 27.7%
  17. " Australian" overpaid defender of Australian security that in Soviet↕ -34.25 (#1) · ⚡ 32.0%
  18. " Australian" Australian Broadcasting Corporation and Australian Stock Exchange and chancellor↕ -34.25 (#1) · ⚡ 32.0%
  19. " Australian" Tooth, a prominent Australian businessman. His brother↕ -34.25 (#1) · ⚡ 30.7%
  20. " Australian" – name changed to Australian Signals Directorate↵↵ASD↕ -34.00 (#1) · ⚡ 31.8%
Gemma 2 2B · L1/C1290

Word related to diagnose

Confidence 5/5
Component profile — C1290 · layer 1 · positive tail

Logit-lens tokens · positive

#1" diagnosis"#2" diagnostic"#3" Diagnosis"#4" diagnose"#5" DIAG"#6" Diagnostic"#7" diagnoses"#8"diagnosis"#9" diagnosed"#10" diagnostics"

R-lens tokens · positive

#1" diagnostic"#2" diagnosis"#3" diagnose"#4" diagnoses"#5" diagnosing"#6" DIAG"#7" diagnos"#8" Diagnosis"#9" Diagnostic"#10" diagnosed"

Skewness

+5.88

Excess kurtosis

220.22rank #1033

Logcosh deviation

0.0978rank #1291

Top-score occurrences · positive

  1. " diagnosis"3 years with a diagnosis of lower urinary tract↕ +44.25 (#1) · ⚡ 44.3%
  2. " diagnosis".7 cm at diagnosis. We report the↕ +44.25 (#1) · ⚡ 43.4%
  3. " diagnosis" of ultrasound in the diagnosis of prostate cancer will↕ +44.00 (#1) · ⚡ 45.9%
  4. " diagnostic"].↵Complications in diagnostic mediastinoscopy↕ +43.50 (#1) · ⚡ 36.7%
  5. " diagnosis" also help in the diagnosis, being the N↕ +43.50 (#1) · ⚡ 47.1%
  6. " diagnosis" Here we discuss the diagnosis and treatment of NC↕ +43.25 (#1) · ⚡ 43.4%
  7. " diagnosis" at the time of diagnosis. The presence of↕ +43.00 (#1) · ⚡ 45.1%
  8. " diagnosis" for the sonographic diagnosis of both malignant and↕ +43.00 (#1) · ⚡ 44.3%
  9. " diagnosis" cells or tissue for diagnosis by the insertion of↕ +43.00 (#1) · ⚡ 43.2%
  10. " diagnosis"8 h with discharge diagnosis codes corresponding to chronic↕ +43.00 (#1) · ⚡ 42.4%
  11. " diagnostic"al resection: a diagnostic challenge.↵Pro↕ +42.75 (#1) · ⚡ 37.7%
  12. " diagnosis"7 children). The diagnosis of CD was based↕ +42.75 (#1) · ⚡ 44.5%
  13. " diagnosis" procedure to achieve a diagnosis, then it would↕ +42.50 (#1) · ⚡ 45.1%
  14. " diagnosis" syndrome (ACS) diagnosis and prognosis. MMP↕ +42.50 (#1) · ⚡ 43.2%
  15. " diagnosis"-up information regarding diagnosis, or were on↕ +42.25 (#1) · ⚡ 44.9%
  16. " diagnostic" pathophysiology, and diagnostic evaluation of EOC↕ +42.00 (#1) · ⚡ 36.9%
  17. " diagnostic" was undertaken. Baseline diagnostic MRI findings included;↕ +41.75 (#1) · ⚡ 37.9%
  18. " diagnosis"-dimer test and diagnosis {#sec2.↕ +41.75 (#1) · ⚡ 44.5%
  19. " diagnosis" analysis strongly supported the diagnosis of a leukemia of↕ +41.75 (#1) · ⚡ 39.5%
  20. " diagnosis" histologically-confirmed diagnosis of PGL were↕ +41.50 (#1) · ⚡ 43.0%
Gemma 2 2B · L1/C2180

Super

Confidence 5/5
Component profile — C2180 · layer 1 · negative tail

Logit-lens tokens · negative

#1" super"#2" Super"#3"super"#4"Super"#5" SUPER"#6"SUPER"#7"uper"#8" супер"#9"超级"#10" siêu"

R-lens tokens · negative

#1" organisation"#2" maximise"#3"organisation"#4"initialise"#5"’."#6" realisation"#7" vapour"#8" neighbour"#9" optimisation"#10" SUPER"

Skewness

-0.20

Excess kurtosis

14.64rank #2050

Logcosh deviation

0.0577rank #2181

Top-score occurrences · negative

  1. " Super"-League and Chinese Super League. The rights↕ -18.50 (#3) · ⚡ 5.2%
  2. "Super" override fun visitSuperTypeList(list:↕ -18.38 (#3) · ⚡ 6.3%
  3. " Super", Corporate Counsel for Supermarkets General Corporation.↕ -18.12 (#3) · ⚡ 5.9%
  4. "Super" album Change↵ "Superpower" (song↕ -17.88 (#3) · ⚡ 7.2%
  5. " super"FBT) or super guarantee (SG).↕ -17.75 (#3) · ⚡ 4.9%
  6. " Super" in the Intrust Super Cup in Australia and↕ -17.50 (#3) · ⚡ 6.2%
  7. "super", tracker↵ super tracker↵ @↕ -17.50 (#4) · ⚡ 4.2%
  8. "super") {↵ super(props);↵↕ -17.50 (#3) · ⚡ 4.2%
  9. " Super"019 PA Super 59↵↵↕ -17.38 (#3) · ⚡ 6.4%
  10. " Super"Category:Albanian Superliga players↕ -17.38 (#3) · ⚡ 5.4%
  11. " super"↵ protected String superExecutionId;↵↕ -17.25 (#3) · ⚡ 5.6%
  12. " Super" video game magazine named Super POWER from 1↕ -17.25 (#3) · ⚡ 6.5%
  13. " Super" Kenner Toys↵ Project Superpowers, a comic↕ -17.12 (#3) · ⚡ 6.6%
  14. " Super" men’s AMA Supercross and Motocross↕ -17.00 (#3) · ⚡ 5.6%
  15. " Super" are going to the Super Bowl, maybe they↕ -16.88 (#3) · ⚡ 5.8%
  16. "super"() {↵ super(JDBCT↕ -16.88 (#3) · ⚡ 5.1%
  17. "Super"↵↵Support Groups↵↵Supervised Visitation and Safe↕ -16.75 (#3) · ⚡ 6.1%
  18. "super" {↵ super.onCreateOptionsMenu(↕ -16.75 (#3) · ⚡ 4.5%
  19. "super") {↵ super(db, INDEX↕ -16.75 (#3) · ⚡ 5.1%
  20. " Super". The A3 Super II has an integrated↕ -16.75 (#3) · ⚡ 5.7%
Gemma 2 2B · L2/C1994

Long underscore separator

Confidence 3/5
Component profile — C1994 · layer 2 · negative tail

Logit-lens tokens · negative

#1"English"#2" defense"#3" English"#4" Defense"#5"Defense"#6" COMPONENT"#7" components"#8" elements"#9" DEFENSE"#10"defense"

R-lens tokens · negative

#1"Defense"#2"defense"#3" Defense"#4" defense"#5" DEFENSE"#6"Neighborhood"#7" viață"#8" computadora"#9"toward"#10" theater"

Skewness

-0.05

Excess kurtosis

8.70rank #2097

Logcosh deviation

0.0556rank #1995

Top-score occurrences · negative

  1. "________________"riam opinion.↵↵_________________________________________________________________↕ -14.88 (#2) · ⚡ 4.0%
  2. "________________" APPELLEE↵↵↵↵↵↵_________________________________↵↵↵↵FROM↕ -14.06 (#3) · ⚡ 3.8%
  3. "________________"Circuit Judges.↵____________________________________↵↵YAN↕ -13.31 (#3) · ⚡ 3.3%
  4. "________________"058↵_______________________________________↵↵Y↕ -13.00 (#3) · ⚡ 3.6%
  5. "________________".↵↵↵ _________________________________↵ ↕ -12.94 (#4) · ⚡ 3.0%
  6. "________________"815↵____________________________________↵JUN↕ -12.81 (#3) · ⚡ 3.6%
  7. "________________".↵↵ _________________________________________/↵↵ ↕ -12.81 (#3) · ⚡ 4.0%
  8. "________________" Judges.↵ _______________________________________↵↵ ↕ -12.69 (#3) · ⚡ 3.1%
  9. "________________"↵TABLE I ______________________________________ Wave-↕ -12.44 (#2) · ⚡ 2.1%
  10. "________________"936↵_______________________________________↵↵Y↕ -12.25 (#4) · ⚡ 3.7%
  11. "________________" consideration:↵↵↵↵↵____________________________________________↵↵↵4↵↕ -12.19 (#3) · ⚡ 3.6%
  12. "________________"048↵____________________________________↵↵↵↵ ↕ -12.19 (#4) · ⚡ 3.5%
  13. "________________" Judges.↵____________________________________↵↵YAN Y↕ -12.12 (#3) · ⚡ 3.7%
  14. " Defense"↵↵First-Team Defense Second-Team Defense↕ -12.06 (#3) · ⚡ 2.8%
  15. " Defense" Defense Fourth-Team Defense DL Vita Vea Washington↕ -12.00 (#3) · ⚡ 2.8%
  16. "________________"tor,↵ ______________________________↵ ↕ -12.00 (#4) · ⚡ 2.8%
  17. "________________"787↵____________________________________↵↵JIN↕ -12.00 (#4) · ⚡ 3.5%
  18. "________________"H CIRCUIT↵ ________________________↵↵ No↕ -11.88 (#3) · ⚡ 2.9%
  19. "________________"142↵_______________________________________↵↵ME↕ -11.81 (#4) · ⚡ 3.5%
  20. "________________"456↵_______________________________________↵↵0↕ -11.75 (#4) · ⚡ 3.5%
Gemma 2 2B · L3/C2037

Substring tab or cat

Confidence 5/5
Component profile — C2037 · layer 3 · negative tail

Logit-lens tokens · negative

#1" cat"#2" Cat"#3" tab"#4" cats"#5"Cat"#6" Cats"#7"Tab"#8" CAT"#9" Tab"#10"Cats"

R-lens tokens · negative

#1" cat"#2" Cat"#3" CAT"#4" cats"#5"Cat"#6" stabilisation"#7"cat"#8"Portail"#9"ValueStyle"#10"PMailer"

Skewness

-0.06

Excess kurtosis

5.18rank #2162

Logcosh deviation

0.0439rank #2038

Top-score occurrences · negative

  1. " Tab"↵Make use of TabControl.Selected↵↕ -11.62 (#3) · ⚡ 3.6%
  2. "cat">↵void ConcatCPU(DeviceBase↕ -11.50 (#3) · ⚡ 1.6%
  3. " tab"div class="row tab-pane Galeria">↕ -11.19 (#4) · ⚡ 2.4%
  4. " Cat" stainless Reinhard Takumi Catback, and a↕ -10.94 (#2) · ⚡ 3.4%
  5. " cat"-mill, crazy cat lady, just ↕ -10.88 (#3) · ⚡ 2.2%
  6. " Tab"="string">Extra Tab</argument>↵↕ -10.75 (#2) · ⚡ 3.5%
  7. " Tab"9> - Tab(3)↵↕ -10.69 (#3) · ⚡ 3.0%
  8. " tab" header text on the tab rather than in the↕ -10.56 (#2) · ⚡ 3.4%
  9. " tab" state data survive a tab close/crash you↕ -10.56 (#3) · ⚡ 3.8%
  10. " tab" changes depending on what tab is selected. to↕ -10.44 (#3) · ⚡ 3.1%
  11. " tab" right click on the tab control itself, as↕ -10.44 (#2) · ⚡ 3.3%
  12. "Tab"dusSalam\AddTab\Block\Product↕ -10.44 (#3) · ⚡ 2.7%
  13. " tab" app starts with a tab bar as its main↕ -10.44 (#3) · ⚡ 2.9%
  14. " Tab" pass this DataSource to Tablix. But if↕ -10.44 (#3) · ⚡ 3.0%
  15. " Tab" Order column right in Tablix at design time↕ -10.31 (#3) · ⚡ 3.3%
  16. "cat"remember to install socat)↵konsole↕ -10.25 (#2) · ⚡ 1.1%
  17. " tab" never saw in console tab.↵↵Image ↕ -10.19 (#3) · ⚡ 3.0%
  18. " Tab" Support via the Support Tab web form link or↕ -10.19 (#3) · ⚡ 2.7%
  19. " tab" // Create the tab control DOM structure...↵↕ -10.12 (#3) · ⚡ 3.3%
  20. " tab" ↵CREATE TABLE tablename (↵..........↕ -10.00 (#4) · ⚡ 2.4%
Gemma 2 2B · L4/C271

Why

Confidence 5/5
Component profile — C271 · layer 4 · positive tail

Logit-lens tokens · positive

#1" why"#2"why"#3"Why"#4" Why"#5" reasons"#6" WHY"#7" reason"#8"WHY"#9" pourquoi"#10"reasons"

R-lens tokens · positive

#1" why"#2" Why"#3"why"#4" WHY"#5"Why"#6" reasons"#7" reason"#8"WHY"#9" Waarom"#10" Reasons"

Skewness

+9.26

Excess kurtosis

251.63rank #168

Logcosh deviation

0.1237rank #272

Top-score occurrences · positive

  1. "qué":↵↵¿Porqué en este loop de↕ +48.00 (#1) · ⚡ 30.9%
  2. "si"ä:↵↵Miksi se näyttää t↕ +47.00 (#1) · ⚡ 33.6%
  3. " qué":↵↵¿Por qué _=$=+[],↕ +43.25 (#1) · ⚡ 34.8%
  4. " y"...↵Cant understand y OutOfMemoryException↕ +41.50 (#1) · ⚡ 33.6%
  5. "ys" chronicling the whys of product successes and↕ +41.00 (#1) · ⚡ 25.2%
  6. " why", I kept wondering why the Yahoo people chose↕ +38.00 (#1) · ⚡ 46.1%
  7. " why" don't know why I keep posting in↕ +36.25 (#1) · ⚡ 36.9%
  8. " why" people wanted to know why it had taken so↕ +35.50 (#1) · ⚡ 47.1%
  9. " why". Many asked me why I was going back↕ +35.50 (#1) · ⚡ 38.5%
  10. " why" and demands to know why nothing is being done↕ +35.50 (#1) · ⚡ 48.2%
  11. " why"- Tell me: why did not you go↕ +35.25 (#1) · ⚡ 37.9%
  12. " why"'t ask me why it was in there↕ +35.00 (#1) · ⚡ 38.9%
  13. " Why"You may wonder: Why the strange phrasing:↕ +34.75 (#1) · ⚡ 44.1%
  14. " why" Hutto. Asked why they have not read↕ +34.75 (#1) · ⚡ 33.2%
  15. " why" I frequently am asked why I’m running↕ +34.50 (#1) · ⚡ 34.2%
  16. " why", and asked him why Tony Roberts, who↕ +34.50 (#1) · ⚡ 36.5%
  17. " why" don't know why I cannot find a↕ +34.25 (#1) · ⚡ 43.9%
  18. " why"'m not sure why you're trying↕ +34.25 (#1) · ⚡ 35.2%
  19. " why" Illinois. As for why this scheme worked:↕ +34.25 (#1) · ⚡ 41.8%
  20. " why". They asked us why we were↵on↕ +34.00 (#1) · ⚡ 36.1%
Gemma 2 2B · L4/C494

Programming errors / exceptions

Confidence 5/5
Component profile — C494 · layer 4 · positive tail

Logit-lens tokens · positive

#1" error"#2" Error"#3"Error"#4"Exception"#5"error"#6" ERROR"#7" Exception"#8"<bos>"#9" errors"#10"ERROR"

R-lens tokens · positive

#1" Error"#2" error"#3" ERROR"#4" Exception"#5"Error"#6" errors"#7"Exception"#8" Errors"#9"error"#10" failure"

Skewness

+5.50

Excess kurtosis

100.71rank #711

Logcosh deviation

0.1045rank #495

Top-score occurrences · positive

  1. "OD" or not and BSOD with "Kernel notification↕ +35.00 (#1) · ⚡ 18.6%
  2. " exception", I get this exception: Invalid token for↕ +35.00 (#1) · ⚡ 38.5%
  3. " exception" than you get the exception as predicted_index↕ +33.50 (#1) · ⚡ 36.9%
  4. "ception"you should catch exeception and show to know↕ +32.25 (#1) · ⚡ 16.3%
  5. "Exception" I get an IllegalArgumentException. I use G↕ +30.50 (#1) · ⚡ 25.6%
  6. " raised""""An error was raised. This is used↕ +29.75 (#1) · ⚡ 21.4%
  7. " exception" got an access violation exception in my code.↕ +29.75 (#1) · ⚡ 27.1%
  8. " error"'m getting an error,↵(Back↕ +29.50 (#1) · ⚡ 27.3%
  9. " new")↵ throw new InvalidDataException($"↕ +29.38 (#1) · ⚡ 16.7%
  10. " error"I am getting this error when I click on↕ +29.25 (#1) · ⚡ 27.3%
  11. " exception" and will cause an exception.↵ *↕ +28.88 (#1) · ⚡ 29.9%
  12. " raised"↵an exception is raised, which list the↕ +28.75 (#1) · ⚡ 19.0%
  13. "tion"tofmemoryexcetion.↵↵In .↕ +28.75 (#1) · ⚡ 17.4%
  14. " new"(T))) throw new Exception();↵ ↕ +28.75 (#2) · ⚡ 16.4%
  15. " new"{↵ throw new NotImplementedException();↵↕ +28.75 (#2) · ⚡ 14.7%
  16. " error"I receive the following error:↵Invalid date↕ +28.62 (#1) · ⚡ 30.3%
  17. " error" I am receiving this error:↵Could not↕ +28.50 (#1) · ⚡ 27.9%
  18. "raise"0:↵ raise inputError("The↕ +28.25 (#1) · ⚡ 12.8%
  19. " error" ends up throwing an error sense it is a↕ +28.12 (#1) · ⚡ 32.0%
  20. " error"↵But gives me error for IE9.↕ +28.12 (#1) · ⚡ 26.8%
Gemma 2 2B · L4/C1229

Letter Q

Confidence 5/5
Component profile — C1229 · layer 4 · negative tail

Logit-lens tokens · negative

#1" Quar"#2" QU"#3" CQ"#4" quiche"#5" QS"#6"qtype"#7" qs"#8" QM"#9" Quay"#10" qc"

R-lens tokens · negative

#1" CQ"#2" QS"#3" AQ"#4" Qt"#5" dq"#6" QU"#7" qs"#8" QM"#9" RQ"#10" qt"

Skewness

-2.36

Excess kurtosis

29.08rank #1688

Logcosh deviation

0.0751rank #1230

Top-score occurrences · negative

  1. "q"‚llo↵<qwebirc32↕ -20.38 (#1) · ⚡ 13.4%
  2. " q":↵command: qsnip hello↵effect↕ -19.00 (#1) · ⚡ 15.9%
  3. " qu" composite system of two qubits), similar approaches↕ -18.88 (#1) · ⚡ 8.7%
  4. "q", Inc. qX086↕ -18.75 (#1) · ⚡ 13.6%
  5. " qu" curles qnd quicle kisses him and↕ -18.38 (#1) · ⚡ 12.0%
  6. "qu") <mail@qutebrowser.org↕ -18.38 (#1) · ⚡ 11.9%
  7. " qu" read a bit about quines , but it↕ -18.25 (#1) · ⚡ 16.8%
  8. " Q"_data() { QTest::addColumn<↕ -18.25 (#1) · ⚡ 7.9%
  9. "q"GetResultAtIndex(q,i);↵↕ -18.12 (#1) · ⚡ 17.4%
  10. "OT"Null(' + QUOTENAME(ErrClass↕ -18.00 (#1) · ⚡ 4.8%
  11. " q"↵False↵Let q(z) =↕ -18.00 (#2) · ⚡ 10.9%
  12. " q" and click "Run qmake" should fix↕ -17.88 (#1) · ⚡ 12.7%
  13. "Q"QT_USE_QSTRINGBUILDER -↕ -17.88 (#1) · ⚡ 14.0%
  14. "qu"82) (quoting Stanley v.↕ -17.88 (#1) · ⚡ 5.9%
  15. "Q"0.0\Q↵tXml -↕ -17.75 (#1) · ⚡ 9.0%
  16. " Q"FromPoint(const QPoint &point)↕ -17.75 (#1) · ⚡ 6.8%
  17. "Q"↵#include <QThread>↵↵int↕ -17.75 (#1) · ⚡ 5.3%
  18. " qu" be available soon on quasimondo.com↕ -17.75 (#1) · ⚡ 14.2%
  19. " q"↵What is the q'th term of↕ -17.75 (#1) · ⚡ 14.0%
  20. " q"95. Let q be 1 +↕ -17.75 (#2) · ⚡ 11.8%
Gemma 2 2B · L4/C1372

Check

Confidence 5/5
Component profile — C1372 · layer 4 · negative tail

Logit-lens tokens · negative

#1" CHECK"#2" check"#3" Check"#4" checking"#5" Checking"#6" checks"#7"CHECK"#8"check"#9" checked"#10" Checks"

R-lens tokens · negative

#1" CHECK"#2" Checking"#3" checking"#4"CHECK"#5" Checker"#6" Check"#7" checker"#8" checks"#9" checkers"#10" check"

Skewness

-3.42

Excess kurtosis

74.02rank #1003

Logcosh deviation

0.0703rank #1373

Top-score occurrences · negative

  1. " check", and mark a check box for the "↕ -32.00 (#1) · ⚡ 34.2%
  2. " check" judge he'd check himself into rehab.↕ -31.38 (#1) · ⚡ 33.8%
  3. "KS", ABS(CHEKSUM(RAND(↕ -30.25 (#1) · ⚡ 17.7%
  4. " check" "box" to check on a postal card↕ -30.12 (#1) · ⚡ 24.8%
  5. " check", make sure to check out these posts:↕ -30.12 (#1) · ⚡ 33.4%
  6. " Checking" playing or recording. Checking this box allows you↕ -29.00 (#1) · ⚡ 31.4%
  7. " Check" and San Marcos. Check out our online behavioral↕ -29.00 (#1) · ⚡ 35.7%
  8. " check". You ought to check it out, but↕ -28.38 (#1) · ⚡ 33.2%
  9. " check" remote database, and check when the last time↕ -28.00 (#1) · ⚡ 38.3%
  10. " Check" are | centered | Check out the documentation for↕ -28.00 (#1) · ⚡ 33.6%
  11. " check" the items that I check out are always the↕ -28.00 (#1) · ⚡ 30.9%
  12. " check" click the logo, check hhe you are↕ -28.00 (#1) · ⚡ 33.6%
  13. "Check"<bos>Check out our new site↕ -27.88 (#1) · ⚡ 29.3%
  14. "Check"<bos>Check out our new site↕ -27.88 (#1) · ⚡ 29.1%
  15. "Check" the article.↵↵Check out the table↵↵↕ -27.88 (#1) · ⚡ 39.5%
  16. "Check"<bos>Check out our new site↕ -27.88 (#1) · ⚡ 29.1%
  17. "Check"<bos>Check out our new site↕ -27.88 (#1) · ⚡ 29.1%
  18. " check" I didn't check it out yet,↕ -27.38 (#1) · ⚡ 30.3%
  19. " check"↵↵Be sure to check out the highlights from↕ -27.38 (#1) · ⚡ 23.8%
  20. "check" a knife or "check out." However,↕ -27.25 (#1) · ⚡ 30.7%
Gemma 2 2B · L4/C2205

Fast

Confidence 5/5
Component profile — C2205 · layer 4 · negative tail

Logit-lens tokens · negative

#1" faſt"#2" FAST"#3"fast"#4"Fast"#5"FAST"#6"rapid"#7" fast"#8"Swift"#9" rapid"#10"Rapid"

R-lens tokens · negative

#1" FAST"#2" rapide"#3"FAST"#4" rapida"#5" Fast"#6"Swift"#7"Fast"#8" Faster"#9"Fastest"#10" Fastest"

Skewness

-0.08

Excess kurtosis

3.62rank #2218

Logcosh deviation

0.0320rank #2206

Top-score occurrences · negative

  1. " Fast"time↵Manufacturing and Fast Delivery services.↵↵↕ -12.88 (#3) · ⚡ 7.2%
  2. " fast"↵competitive price, fast delivery and top services↕ -12.62 (#3) · ⚡ 6.0%
  3. " fast" Product Express - a fast, easy way to↕ -12.06 (#3) · ⚡ 6.0%
  4. " fast" erectable configuration for fast field installation; a↕ -12.00 (#3) · ⚡ 5.5%
  5. " fast" for cable television and fast internet speeds as top↕ -11.69 (#3) · ⚡ 5.6%
  6. " fast" want to do this fast and effectively. In↕ -11.50 (#4) · ⚡ 5.2%
  7. "Fast"00d↵↵Fastens with faste↕ -11.38 (#3) · ⚡ 5.2%
  8. "fast" SDK is super-fast customizable Full text Auto↕ -11.00 (#3) · ⚡ 3.5%
  9. " rapid" powerful tool for the rapid construction of substituted tetra↕ -10.88 (#3) · ⚡ 3.9%
  10. " fast"photon fluorescence microscope with fast-scanning capabilities and↕ -10.75 (#3) · ⚡ 3.9%
  11. " fast" Option valuation using the fast Fourier transform↵by↕ -10.38 (#3) · ⚡ 4.8%
  12. " rapid" new strategy for a rapid identification of heteroplasmic↕ -10.31 (#3) · ⚡ 3.8%
  13. " rapid" configuration files that allow rapid switching between Flows.↕ -10.25 (#3) · ⚡ 3.7%
  14. " fast", also known as fast-track surgery,↕ -10.19 (#3) · ⚡ 5.1%
  15. " fast" environment which may be fast-paced or changing↕ -10.19 (#3) · ⚡ 4.6%
  16. "Fast" every day.↵↵Fast forward a few years↕ -10.12 (#3) · ⚡ 4.9%
  17. " Fast"'s New England Fast 50↵↵↕ -10.12 (#3) · ⚡ 3.8%
  18. " fast" It's a fast-paced, dramatic↕ -10.06 (#3) · ⚡ 3.7%
  19. " Fast" A Simple, Fast Dominance Algorithm.↕ -10.00 (#3) · ⚡ 3.1%
  20. " fast" like my query string fast to match all of↕ -9.94 (#3) · ⚡ 3.6%
Gemma 2 2B · L5/C1839

U.S. Supreme Court case citation

Confidence 4/5
Component profile — C1839 · layer 5 · positive tail

Logit-lens tokens · positive

#1"or"#2":✨"#3"S"#4"M"#5"k"#6" or"#7":"#8"ri"#9" "#10"y"

R-lens tokens · positive

#1" MenuView"#2" "#3"styleable"#4"</i>"#5" + "#6"displayquote"#7" Anam"#8" Magist"#9" Wadi"#10" < "

Skewness

+3.46

Excess kurtosis

181.31rank #145

Logcosh deviation

0.0422rank #1840

Top-score occurrences · positive

  1. "S"96 U.S. 82↕ +59.75 (#1) · ⚡ 33.8%
  2. "S"36 U.S. 65↕ +59.50 (#1) · ⚡ 35.9%
  3. "S"41 U.S. 91↕ +59.00 (#1) · ⚡ 33.6%
  4. "S"12 U.S. 47↕ +57.50 (#1) · ⚡ 30.5%
  5. "S"56 U.S. 64↕ +57.00 (#1) · ⚡ 34.6%
  6. "S"39 U.S. 55↕ +56.75 (#1) · ⚡ 33.2%
  7. "S"86 U.S. 73↕ +56.25 (#1) · ⚡ 35.9%
  8. "S"82 U.S. 25↕ +56.00 (#1) · ⚡ 34.6%
  9. "S"54 U.S. 88↕ +55.75 (#1) · ⚡ 32.0%
  10. "S"90 U.S. 43↕ +55.50 (#1) · ⚡ 30.1%
  11. "S"81 U.S. 93↕ +55.00 (#1) · ⚡ 34.2%
  12. "S"16 U.S. 94↕ +54.50 (#1) · ⚡ 30.1%
  13. "S"61 U.S. 21↕ +54.25 (#1) · ⚡ 33.8%
  14. "S"65 U.S. 16↕ +53.75 (#1) · ⚡ 32.2%
  15. "S"9↵U.S. 49↕ +53.50 (#1) · ⚡ 30.7%
  16. "S"84 U.S. 43↕ +52.75 (#1) · ⚡ 33.2%
  17. "S"46 U.S. 35↕ +51.50 (#1) · ⚡ 28.7%
  18. "S"59 U.S. 99↕ +51.50 (#1) · ⚡ 29.9%
  19. "S", ___ U.S. ___, ↕ +35.00 (#1) · ⚡ 16.3%
  20. "S", --- U.S. ----, ↕ +33.50 (#1) · ⚡ 15.7%
Gemma 2 2B · L6/C1503

Definitional statement

Confidence 5/5
Component profile — C1503 · layer 6 · negative tail

Logit-lens tokens · negative

#1"*/;"#2"};*/"#3"';\r"#4" <\","#5"\\\",\\"#6")*/"#7"]:\r"#8"')\";"#9"\";\r"#10"\"</"

R-lens tokens · negative

#1" individuals"#2" commonly"#3" information"#4" organisms"#5" occurs"#6" known"#7" technology"#8" forms"#9" society"#10" language"

Skewness

-2.15

Excess kurtosis

22.26rank #1606

Logcosh deviation

0.0566rank #1504

Top-score occurrences · negative

  1. " use" it is “the use of digital information and↕ -27.88 (#1) · ⚡ 22.2%
  2. " who" alibi witness (someone who can provide an account↕ -26.88 (#1) · ⚡ 24.6%
  3. " promotion"Gambling advertising is the promotion of gambling by casinos↕ -25.88 (#1) · ⚡ 18.2%
  4. " instance"↵↵a state or instance of serious or continued↕ -24.38 (#1) · ⚡ 10.4%
  5. " of" is “the use of digital information and communication↕ -23.62 (#1) · ⚡ 23.5%
  6. " use" defined as 'the use of current best evidence↕ -22.88 (#1) · ⚡ 19.7%
  7. " of" is the complete replacement of a damaged hip with↕ -22.25 (#1) · ⚡ 20.5%
  8. " movement" is "any bodily movement produced by skeletal muscles↕ -21.50 (#1) · ⚡ 10.6%
  9. " which" a biochemical condition in which oxidant species overwhelm↕ -21.25 (#1) · ⚡ 18.9%
  10. " creating" is the procedure of creating software applications which follow↕ -20.62 (#1) · ⚡ 17.5%
  11. " that"fiable information is anything that is unique to a↕ -20.62 (#1) · ⚡ 18.1%
  12. " which" event as one in which at least four people↕ -20.25 (#1) · ⚡ 14.0%
  13. " that" - these are proteins that are present inside the↕ -20.00 (#1) · ⚡ 20.8%
  14. " use"int") is the use of internal radionuclides↕ -20.00 (#1) · ⚡ 14.5%
  15. " ability" mental health is the ability to establish harmonious communication↕ -19.75 (#1) · ⚡ 14.0%
  16. " piece" A wedge-shaped piece of cloth taken out↕ -19.50 (#1) · ⚡ 9.9%
  17. " of" automatic re-building of project code base that↕ -19.00 (#1) · ⚡ 15.6%
  18. "uette"figurine   a statuette, especially one of↕ -19.00 (#1) · ⚡ 8.2%
  19. " which" hardware is hardware for which the design is made↕ -18.88 (#2) · ⚡ 10.5%
  20. " of"olysis is the breakdown of skeletal muscle that can↕ -18.88 (#1) · ⚡ 16.0%
Gemma 2 2B · L6/C2082

?

Confidence 0/5
Component profile — C2082 · layer 6 · positive tail

Logit-lens tokens · positive

#1" myſelf"#2" Theſe"#3" purpoſe"#4" itſelf"#5"ſelf"#6" uſe"#7" houſe"#8" ſeveral"#9" Majefty"#10" Chriftian"

R-lens tokens · positive

#1" poética"#2"…)."#3" kath"#4" …”"#5" Argyle"#6"…\""#7"…”"#8"...\"."#9"Prague"#10" fornic"

Skewness

+0.06

Excess kurtosis

2.45rank #2193

Logcosh deviation

0.0273rank #2083

Top-score occurrences · positive

  1. " Dance" You Think You Can Dance” and new show↕ +10.12 (#4) · ⚡ 1.1%
  2. "autorest"Azure/go-autorest/autorest"↵↕ +9.50 (#6) · ⚡ 0.4%
  3. " push"#pragma clang diagnostic push↵#pragma clang↕ +9.25 (#5) · ⚡ 0.7%
  4. "Map" new OpenLayers.Map('map');↵↕ +9.00 (#5) · ⚡ 0.9%
  5. " diaper" been through the cloth diaper routine a few times↕ +9.00 (#5) · ⚡ 1.2%
  6. "Map" new OpenLayers.Map('map');↵↕ +9.00 (#3) · ⚡ 0.9%
  7. " ee"↵ reducer: ee.Reducer.count↕ +8.88 (#10) · ⚡ 0.6%
  8. " therm" the summer season the thermals can start popping↕ +8.44 (#9) · ⚡ 0.8%
  9. " wedding").↵↵Planning a wedding can be stressful,↕ +8.38 (#5) · ⚡ 1.4%
  10. "​​​​" effect of, ​​​​​​“You killed people↕ +8.25 (#4) · ⚡ 0.7%
  11. "RED" Glossary of Node-RED terms↵↵For those↕ +8.25 (#5) · ⚡ 0.7%
  12. "pace"package org.dspace.servicemanager↕ +8.25 (#11) · ⚡ 0.4%
  13. "apimachinery"s.io/apimachinery/pkg/apis↕ +8.06 (#22) · ⚡ 0.3%
  14. "links"book/addon-links/register' ;↕ +7.97 (#4) · ⚡ 0.6%
  15. "query"global $wp_query;↵ the↕ +7.94 (#5) · ⚡ 0.6%
  16. " Weddings" a Place for Destination Weddings↵↵Getting married abroad↕ +7.88 (#4) · ⚡ 1.2%
  17. " nursing" for themselves and other nursing colleagues.↕ +7.88 (#5) · ⚡ 1.4%
  18. "encrypt"=/etc/letsencrypt/live/example↕ +7.81 (#10) · ⚡ 0.5%
  19. "Scene" javafx.scene.Scene;↵import javafx↕ +7.81 (#7) · ⚡ 0.4%
  20. " Echo" included all-new Echo Dot, Fire TV↕ +7.81 (#9) · ⚡ 0.9%
Gemma 2 2B · L7/C39

Post-Soviet context

Confidence 5/5
Component profile — C39 · layer 7 · positive tail

Logit-lens tokens · positive

#1"Билгалдахарш"#2"onViewCreated"#3"с"#4"//"#5"#"#6" Moreover"#7" dispu"#8" «"#9"expandindo"#10"Moreover"

R-lens tokens · positive

#1" russian"#2" Russian"#3" Ukrainian"#4" peculiarities"#5" soviet"#6" advices"#7"russian"#8" Yandex"#9" Bulgarian"#10" Soviet"

Skewness

+6.79

Excess kurtosis

84.98rank #485

Logcosh deviation

0.1568rank #40

Top-score occurrences · positive

  1. " State" St-Petersburg State Polytechnic University, ↕ +27.12 (#2) · ⚡ 13.2%
  2. "ly"var R. Mavlyutov'↵↕ +26.25 (#1) · ⚡ 14.6%
  3. " well" interventricular septum as well between COPD patients and↕ +26.00 (#1) · ⚡ 28.7%
  4. "inova" of Natalya Savinova showcases works, which↕ +25.88 (#1) · ⚡ 12.5%
  5. " aim" the disease. The aim: The main goal↕ +25.50 (#1) · ⚡ 16.8%
  6. " Now" doesn't. Now I want to separate↕ +25.38 (#1) · ⚡ 21.8%
  7. " revealed" reaction)-analysis has revealed the presence of viral↕ +25.25 (#1) · ⚡ 22.9%
  8. "hev"Slava Imeshev is president of View↕ +25.12 (#1) · ⚡ 16.7%
  9. " important", could play an important role in development of↕ +25.00 (#1) · ⚡ 19.1%
  10. " rats" outbred white male rats. The gained results↕ +25.00 (#1) · ⚡ 17.1%
  11. " consideration" individual therapeutic program with consideration for the risk group↕ +24.38 (#1) · ⚡ 16.9%
  12. " background" period, against the background of linear increase in↕ +23.75 (#1) · ⚡ 12.4%
  13. " break" a nuclear conflict may break out already in the↕ +23.62 (#1) · ⚡ 16.4%
  14. "'" was checked by Doppler's pulsating focused↕ +23.62 (#1) · ⚡ 31.1%
  15. " did"-6 hrs and did not alter at the↕ +23.50 (#1) · ⚡ 21.3%
  16. " The"0 mm Hg. The blood perfusion in the↕ +23.38 (#1) · ⚡ 31.6%
  17. "’". These measures don’t↵↵foresee↕ +23.25 (#1) · ⚡ 27.1%
  18. "odo".com/feodosiy Views: ↕ +23.25 (#1) · ⚡ 13.0%
  19. "kop" Rural Settlement of Maykopsky District, Russia↕ +23.25 (#1) · ⚡ 11.6%
  20. "uto" R. Mavlyutov'↵title↕ +23.25 (#1) · ⚡ 13.1%
Gemma 2 2B · L7/C878

Code syntax class="

Confidence 5/5
Component profile — C878 · layer 7 · negative tail

Logit-lens tokens · negative

#1" class"#2" Class"#3"classnames"#4" CLASS"#5"Class"#6"=\""#7"class"#8"CLASS"#9"addClass"#10" classes"

R-lens tokens · negative

#1"classnames"#2"Class"#3"addClass"#4"CLASS"#5" avoient"#6" CLASS"#7" Class"#8"className"#9"class"#10"hasClass"

Skewness

-5.40

Excess kurtosis

143.51rank #151

Logcosh deviation

0.0717rank #879

Top-score occurrences · negative

  1. "=\""="button" class="navbar-toggle"↕ -34.25 (#1) · ⚡ 37.7%
  2. "=\""<bos><div class="subblock">↵↕ -34.25 (#1) · ⚡ 32.6%
  3. "=\""↵<div class="wrapper">↵ ↕ -34.00 (#1) · ⚡ 27.0%
  4. "=\""<bos><section class="notice-previous">↕ -33.75 (#1) · ⚡ 27.1%
  5. "=\"" <div class="container">↵ ↕ -33.25 (#1) · ⚡ 35.5%
  6. "=\"" <form class="navbar-form navbar↕ -33.25 (#1) · ⚡ 30.1%
  7. " class"<bos><div class="subblock">↕ -33.00 (#1) · ⚡ 23.6%
  8. "=\"" <p class="lead">抱歉！↕ -32.50 (#1) · ⚡ 37.3%
  9. "=\"" <div class="navbar-header">↵↕ -32.50 (#1) · ⚡ 35.0%
  10. "=\"" <div class="container-fluid">↕ -32.25 (#1) · ⚡ 32.4%
  11. "=\""↵<footer class="page-footer transparent↕ -32.25 (#1) · ⚡ 29.7%
  12. "=\"" <div class="card-header">↕ -32.00 (#1) · ⚡ 31.4%
  13. "=\"" <a class="btn btn-default↕ -32.00 (#1) · ⚡ 39.3%
  14. "=\"" <div class="content-header">↕ -32.00 (#1) · ⚡ 28.5%
  15. "=\""↵↵<div class="demos-nav">↕ -32.00 (#1) · ⚡ 32.4%
  16. "=\""="submit" class="btn btn-default↕ -31.75 (#1) · ⚡ 35.9%
  17. "=\"" Wishlist')) ?>" class="button"><span><↕ -31.75 (#1) · ⚡ 33.2%
  18. "=\""↵<div class="fixedNav">↵↕ -31.62 (#1) · ⚡ 30.7%
  19. "=\""<h4 class="panel-title">↕ -31.62 (#1) · ⚡ 33.8%
  20. "=\"" <ul class="tablist">↵↕ -31.50 (#1) · ⚡ 28.9%
Gemma 2 2B · L7/C1016

Because of, a cause/reason expression

Confidence 5/5
Component profile — C1016 · layer 7 · positive tail

Logit-lens tokens · positive

#1" the"#2"__(/*!"#3"AndEndTag"#4"AddTagHelper"#5" بسبب"#6"ůli"#7" nhờ"#8" adanya"#9"rrggbb"#10"ftagPool"

R-lens tokens · positive

#1" technological"#2" –"#3" numerous"#4" “"#5" injuries"#6" unforeseen"#7" various"#8" unfavorable"#9" ‘"#10" injury"

Skewness

+2.61

Excess kurtosis

28.14rank #1412

Logcosh deviation

0.0680rank #1017

Top-score occurrences · positive

  1. " of" have unintended consequences because of their effects on incentives↕ +19.62 (#1) · ⚡ 19.6%
  2. " of" to divorce court because of cats. Yes!↕ +19.38 (#1) · ⚡ 16.8%
  3. " from"td1 might result from the special protein conformation↕ +19.25 (#1) · ⚡ 17.1%
  4. " of" wrinkles often occur because of the way your muscles↕ +19.25 (#1) · ⚡ 18.8%
  5. " of" ice formation occurs because of the low ambient temperature↕ +18.88 (#1) · ⚡ 20.1%
  6. " to" at Monza owed much to Bottas disturbing Kimi↕ +18.75 (#1) · ⚡ 13.5%
  7. " from" decrease in sensitivity resulted from reduced intracellular doxor↕ +18.38 (#1) · ⚡ 16.5%
  8. " of" that exists only because of the support of our↕ +18.25 (#1) · ⚡ 11.3%
  9. " due"4.↵↵Partly due to the influx of↕ +18.12 (#1) · ⚡ 8.1%
  10. " of" terminated one pregnancy because of genetic abnormalities. This↕ +18.12 (#1) · ⚡ 17.9%
  11. " of" And better developed because of the collaboration of many↕ +18.12 (#1) · ⚡ 16.6%
  12. " of" asphyxia because of neck compression.↵↵↕ +18.12 (#1) · ⚡ 17.9%
  13. " from" has in part resulted from the best, centrally↕ +18.12 (#1) · ⚡ 14.8%
  14. " of" alertness is the result of rapid development of feed↕ +18.00 (#1) · ⚡ 20.2%
  15. " to" are due in part to partial blockages of↕ +18.00 (#1) · ⚡ 19.4%
  16. " from"Pt/Film results from the unique flexibility of↕ +18.00 (#1) · ⚡ 13.8%
  17. " of" insight and data because of the LinkedIn members that↕ +17.88 (#1) · ⚡ 15.8%
  18. " to" owe their staying power to a visionary leader –↕ +17.75 (#1) · ⚡ 12.4%
  19. " of"<bos>Because of the extreme temperatures of↕ +17.50 (#2) · ⚡ 7.5%
  20. " of" induced as a result of the pacing will typically↕ +17.38 (#1) · ⚡ 20.0%
Gemma 2 2B · L8/C138

Previously introduced unfamiliar proper noun

Confidence 4/5
Component profile — C138 · layer 8 · negative tail

Logit-lens tokens · negative

#1" '<?"#2"ENEFITS"#3" '{@"#4" lepto"#5"ERÍA"#6"orszá"#7"UBLIC"#8"ſu"#9"ſo"#10" biographer"

R-lens tokens · negative

#1" démocr"#2" prochaines"#3" mères"#4" enfans"#5" pertes"#6" guerres"#7" normaux"#8" dieux"#9" détru"#10" feroit"

Skewness

-3.96

Excess kurtosis

29.91rank #1364

Logcosh deviation

0.1187rank #139

Top-score occurrences · negative

  1. " Trou" main thoroughfare in Trousdale Estates.↕ -14.50 (#1) · ⚡ 8.6%
  2. "Gib" the lecture.↵↵Gibran Chair Director May↕ -13.81 (#1) · ⚡ 7.3%
  3. " Kao": Who is the Kaotik poet? How↕ -13.62 (#1) · ⚡ 5.8%
  4. "UK"sivism.↵↵UKIP’s appeals↕ -13.38 (#1) · ⚡ 7.3%
  5. " DAY"2017 DAYTONA 5↕ -13.31 (#1) · ⚡ 6.4%
  6. " Nok" line XML file using Nokogiri.↵↕ -13.12 (#1) · ⚡ 5.5%
  7. " Spl" to co-produce Splinter Cell, though↕ -13.00 (#1) · ⚡ 5.4%
  8. " stan" invention to provide a stanchion holder that↕ -12.81 (#1) · ⚡ 8.4%
  9. "cha" (1) [*chameleon* ]{}↕ -12.75 (#1) · ⚡ 6.3%
  10. " ch"3).↵A cholesteric liquid crystal↕ -12.75 (#1) · ⚡ 8.6%
  11. "Flo" helper bacteria.↵Floccularia luteov↕ -12.69 (#1) · ⚡ 7.2%
  12. "Dorf" in the Box↵↵Dorfman Pacific Men'↕ -12.69 (#2) · ⚡ 5.0%
  13. " hom" important to examine the homoscedasticity.↕ -12.62 (#1) · ⚡ 7.0%
  14. " apo"), a marker for apocrine differentiation.↕ -12.56 (#1) · ⚡ 7.1%
  15. " Khu"00 BC, Khuzestan was the↕ -12.56 (#1) · ⚡ 6.5%
  16. " Mud"0 acres. The Mudanjiang City Mega↕ -12.56 (#1) · ⚡ 6.9%
  17. " fos" ring is required in fosinopril sodium↕ -12.50 (#1) · ⚡ 7.5%
  18. " Bu" a volunteer at the Bulosan Center.↕ -12.44 (#1) · ⚡ 7.7%
  19. " Deliver" rate.↵↵The Deliveroo affair follows on↕ -12.44 (#1) · ⚡ 6.2%
  20. " Polar" of the sale of Polarcus Samur -↕ -12.44 (#1) · ⚡ 6.1%
Gemma 2 2B · L10/C834

Website/E-mail domain name

Confidence 5/5
Component profile — C834 · layer 10 · negative tail

Logit-lens tokens · negative

#1"enumi"#2"fi"#3"pa"#4"fl"#5" lietu"#6" transférez"#7" нем"#8"now"#9" tubercle"#10"ity"

R-lens tokens · negative

#1"@"#2"online"#3"dont"#4"now"#5"web"#6" république"#7"we"#8" dna"#9" irish"#10"you"

Skewness

-2.38

Excess kurtosis

27.79rank #904

Logcosh deviation

0.0599rank #835

Top-score occurrences · negative

  1. "star"↵www.energystar.gov/index↕ -19.38 (#1) · ⚡ 7.3%
  2. "l"://davidgrudl.com)↵↕ -19.12 (#1) · ⚡ 12.5%
  3. "min"://www.carmineappice.net↕ -18.88 (#1) · ⚡ 8.6%
  4. "."ho@oec.uni-osna↕ -18.75 (#1) · ⚡ 16.3%
  5. "pages"@support.carepages.com.↵↵↕ -18.62 (#1) · ⚡ 9.0%
  6. "roup"@thehapgroup.com.↕ -18.38 (#1) · ⚡ 10.4%
  7. "."son@lynx.let.hokud↕ -18.12 (#1) · ⚡ 13.2%
  8. "." @ thesurvivalpodcast.com with “question↕ -17.88 (#1) · ⚡ 15.6%
  9. "bin"emberjs.jsbin.com/iz↕ -17.75 (#1) · ⚡ 5.6%
  10. "ings" visit www.billingsfarm.org/↕ -17.62 (#1) · ⚡ 10.0%
  11. "atik"://oxygen.informatik.tu-cott↕ -17.38 (#1) · ⚡ 8.8%
  12. "gothic"https://reimagininggothic.com ).↕ -17.38 (#1) · ⚡ 9.3%
  13. "chron", to photos@chron.com and we↕ -17.38 (#2) · ⚡ 6.3%
  14. "long"http://www.longbeachfilmfestival.↕ -17.38 (#1) · ⚡ 15.6%
  15. "bot"www.gtalbot.org/Browser↕ -17.25 (#1) · ⚡ 9.2%
  16. "tier"://www.viewtier.com, makers↕ -17.25 (#1) · ⚡ 9.7%
  17. "."://textsniper.app/↵======↕ -17.12 (#1) · ⚡ 16.9%
  18. "."@tsinghua.edu.cn,↕ -17.12 (#1) · ⚡ 15.6%
  19. "."://gracefulatheist.wordpress.com/↕ -17.12 (#1) · ⚡ 13.5%
  20. "."ewyearinhawaii.com.↕ -17.00 (#1) · ⚡ 14.3%
Gemma 2 2B · L10/C2197

?

Confidence 0/5
Component profile — C2197 · layer 10 · positive tail

Logit-lens tokens · positive

#1"sigs"#2"dle"#3"wikidata"#4"hy"#5"切"#6"息"#7" ph"#8" schne"#9"ju"#10" pere"

R-lens tokens · positive

#1" stupid"#2" stupidly"#3"Stupid"#4" fucking"#5" Stupid"#6"stupid"#7"Cheap"#8" cheaply"#9" FUCKING"#10" desesper"

Skewness

+0.00

Excess kurtosis

0.84rank #2232

Logcosh deviation

0.0112rank #2198

Top-score occurrences · positive

  1. "rj"OrQ6VrjIEaFf/↕ +6.53 (#5) · ⚡ 0.4%
  2. "ocular" regimens of alternating monocular deprivation. 2↕ +6.47 (#5) · ⚡ 0.9%
  3. " award" the rules for the award recipients:↵↵1↕ +6.44 (#9) · ⚡ 0.6%
  4. "Info"↵ + CategoryInfo : NotSpecified↕ +6.06 (#42) · ⚡ 0.3%
  5. " Tis" dress created by Riccardo Tisci In Pictures:↕ +5.97 (#7) · ⚡ 0.7%
  6. " hint" little more of a hint. You only↕ +5.91 (#7) · ⚡ 0.7%
  7. " open" written as intersection of open subsets of $X↕ +5.88 (#4) · ⚡ 0.7%
  8. "ments" floors and two basements, divided into half↕ +5.84 (#8) · ⚡ 0.6%
  9. " At"↵1.1 At a specified distance:↕ +5.84 (#5) · ⚡ 0.8%
  10. " affiliates" and/or its affiliates. All rights reserved↕ +5.78 (#7) · ⚡ 0.7%
  11. " defense") amongst all freshman defenseman…Was named↕ +5.75 (#6) · ⚡ 0.8%
  12. " VR" Everything about getting into VR is made simpler now↕ +5.72 (#7) · ⚡ 0.9%
  13. "SEND"Intent.ACTION_SEND);↵intent.↕ +5.69 (#19) · ⚡ 0.4%
  14. " safety" health by assuring the safety, effectiveness, and↕ +5.69 (#10) · ⚡ 0.7%
  15. " way"↵ stealthy way:↵↵Various directional↕ +5.66 (#7) · ⚡ 0.8%
  16. " Award" me the Versatile Blogger Award! I’m↕ +5.62 (#31) · ⚡ 0.3%
  17. ":"↵↵Schedule↵Key:↵↵Regular season↵↵↕ +5.59 (#12) · ⚡ 0.6%
  18. "Bezier" maskPath = UIBezierPath(ovalIn↕ +5.59 (#12) · ⚡ 0.4%
  19. "summary"↵↵ /// <summary>↵ ///↕ +5.56 (#13) · ⚡ 0.5%
  20. " affiliates" and/or its affiliates. All rights reserved↕ +5.56 (#8) · ⚡ 0.6%
Gemma 2 2B · L12/C790

Month-day in date

Confidence 4/5
Component profile — C790 · layer 12 · negative tail

Logit-lens tokens · negative

#1"новниш"#2" Púb"#3" Dernière"#4"9"#5"th"#6" dérou"#7" getItemCount"#8"VolleyError"#9"6"#10"Pog"

R-lens tokens · negative

#1"day"#2"ynchronously"#3" Day"#4" day"#5"Day"#6" Friday"#7" Monday"#8" Wednesday"#9" Thursday"#10"9"

Skewness

-2.67

Excess kurtosis

36.86rank #548

Logcosh deviation

0.0580rank #791

Top-score occurrences · negative

  1. "7", January 27, 20↕ -22.50 (#1) · ⚡ 26.4%
  2. "9"Wednesday, September 9, 20↕ -21.38 (#1) · ⚡ 27.0%
  3. "3", June 13, 20↕ -21.12 (#1) · ⚡ 24.1%
  4. "9", September 29, 20↕ -21.12 (#1) · ⚡ 23.9%
  5. "8"Friday, May 8, 20↕ -21.12 (#1) · ⚡ 20.8%
  6. "4", May 24, 20↕ -21.00 (#1) · ⚡ 24.0%
  7. "8"<bos>July 28, 20↕ -20.88 (#1) · ⚡ 21.1%
  8. "6"<bos>February 26 in LGBTQ History↵↵↕ -20.88 (#1) · ⚡ 22.1%
  9. "3"Sunday, April 3, 20↕ -20.88 (#1) · ⚡ 25.0%
  10. "0", September 30, 20↕ -20.75 (#1) · ⚡ 19.9%
  11. "5", June 25, 20↕ -20.75 (#1) · ⚡ 23.6%
  12. "3"Tuesday, April 30, 2↕ -20.62 (#1) · ⚡ 25.0%
  13. "7", February 27, 20↕ -20.50 (#1) · ⚡ 24.2%
  14. "6", May 16, 20↕ -20.25 (#1) · ⚡ 26.0%
  15. "8", February 18, 20↕ -20.25 (#1) · ⚡ 23.4%
  16. "1", July 31, 20↕ -20.12 (#1) · ⚡ 17.8%
  17. "7"Tuesday, July 7, 20↕ -20.00 (#1) · ⚡ 24.6%
  18. "0", November 30, 20↕ -19.88 (#1) · ⚡ 17.1%
  19. "0", March 30, 20↕ -19.75 (#1) · ⚡ 23.6%
  20. "6"Wednesday, April 6, 20↕ -19.75 (#1) · ⚡ 23.9%
Gemma 2 2B · L12/C1615

PHP web-development

Confidence 5/5
Component profile — C1615 · layer 12 · positive tail

Logit-lens tokens · positive

#1" php"#2"php"#3" PHP"#4"PHP"#5" Reſ"#6" ſche"#7" ſch"#8"Php"#9" ſtate"#10" pleaſure"

R-lens tokens · positive

#1" PHP"#2"PHP"#3" php"#4"php"#5"Php"#6" WordPress"#7" Php"#8" <?"#9"WordPress"#10" MySQL"

Skewness

+0.99

Excess kurtosis

9.38rank #1599

Logcosh deviation

0.0332rank #1616

Top-score occurrences · positive

  1. "php"<bos><?php↵$lang['↕ +22.38 (#1) · ⚡ 10.4%
  2. "php"<bos><?php↵/*↵ *↕ +22.38 (#1) · ⚡ 10.4%
  3. "php"<bos><?php↵↵/*↵ *↕ +22.38 (#1) · ⚡ 10.4%
  4. "php"<bos><?php↵// +----------------↕ +22.38 (#1) · ⚡ 10.4%
  5. " subject" This source file is subject to the Open Software↕ +19.75 (#1) · ⚡ 4.4%
  6. "Commerce"↵ use osCommerce\OM\Core↕ +18.50 (#2) · ⚡ 3.9%
  7. " ©"/**↵ * Copyright © Magento, Inc.↕ +17.88 (#1) · ⚡ 3.0%
  8. " Mage"↵ */↵class Mage_Api_Model↕ +17.25 (#2) · ⚡ 6.1%
  9. " Magento"↵/**↵ * Magento↵ *↵ *↕ +16.62 (#2) · ⚡ 5.2%
  10. "Mage"getResponse()->setBody(Mage::helper('core↕ +16.38 (#2) · ⚡ 7.3%
  11. " Mage" $status = Mage::getModel('sales↕ +16.38 (#2) · ⚡ 6.6%
  12. "C" @copyright Copyright (C) 20↕ +15.88 (#3) · ⚡ 2.9%
  13. "Mage"once('app/Mage.php'); ↕ +15.75 (#2) · ⚡ 4.0%
  14. "Mage" * @category Mage↵ * @package↕ +15.56 (#1) · ⚡ 6.2%
  15. " Magento" you wish to upgrade Magento to newer↵ *↕ +15.44 (#1) · ⚡ 5.8%
  16. "Model" Mage_Sales_Model_Order_Config↕ +15.44 (#1) · ⚡ 5.8%
  17. "class">↵ */↵class Mage_Sales_↕ +15.38 (#2) · ⚡ 6.0%
  18. " Mage"', '<?php echo Mage::helper('core↕ +15.25 (#2) · ⚡ 7.0%
  19. " defined"<bos><?php defined('BX_DOL↕ +15.25 (#2) · ⚡ 2.3%
  20. "/*"<bos><?php↵/*↵ * THIS SOFTWARE↕ +15.12 (#3) · ⚡ 3.6%
Gemma 2 2B · L12/C2205

Digit in document-initial Federal Reporter citation

Confidence 2/5
Component profile — C2205 · layer 12 · negative tail

Logit-lens tokens · negative

#1"Trades"#2" sessionFactory"#3"taz"#4" yore"#5" Fawcett"#6" getchar"#7"ξης"#8"fated"#9"iton"#10"כך"

R-lens tokens · negative

#1" piú"#2" modalités"#3" récentes"#4" récents"#5"berdayakan"#6" arranque"#7" gezet"#8" coordon"#9" découver"#10" Fawcett"

Skewness

-0.01

Excess kurtosis

1.02rank #2062

Logcosh deviation

0.0100rank #2206

Top-score occurrences · negative

  1. "4"<bos>74 F.3d↕ -11.88 (#3) · ⚡ 5.2%
  2. "5"<bos>956 F.2↕ -11.62 (#2) · ⚡ 4.5%
  3. "8"<bos>788 F.2↕ -11.25 (#3) · ⚡ 4.8%
  4. "3"<bos>734 F.2↕ -11.19 (#3) · ⚡ 4.5%
  5. "7"<bos>875 F.2↕ -11.00 (#3) · ⚡ 4.5%
  6. "8"<bos>08-152↕ -10.56 (#3) · ⚡ 3.4%
  7. "1"<bos>812 F.2↕ -10.38 (#4) · ⚡ 4.2%
  8. "5"<bos>758 F.2↕ -10.31 (#3) · ⚡ 4.8%
  9. "3"<bos>535 F.2↕ -10.31 (#4) · ⚡ 4.3%
  10. "3"<bos>03)?↵-6↕ -10.25 (#3) · ⚡ 2.9%
  11. "5"<bos>759 F.2↕ -10.25 (#3) · ⚡ 4.7%
  12. "1"<bos>619 F.2↕ -9.94 (#4) · ⚡ 3.8%
  13. "1"<bos>915 F.2↕ -9.94 (#2) · ⚡ 4.1%
  14. "6"<bos>176 page of collection of↕ -9.44 (#3) · ⚡ 4.2%
  15. "6"<bos>1761 in Denmark↵↵↕ -9.38 (#3) · ⚡ 4.1%
  16. "6"<bos>162 F.3↕ -8.88 (#4) · ⚡ 3.8%
  17. "8"<bos>183 F.3↕ -8.88 (#5) · ⚡ 3.9%
  18. "9"<bos>1912 Müref↕ -8.75 (#3) · ⚡ 3.4%
  19. "9"<bos>1942 Iowa Pre↕ -8.75 (#3) · ⚡ 3.4%
  20. "6"<bos>1265 in Scotland↵↵↕ -8.62 (#3) · ⚡ 3.8%
Gemma 2 2B · L12/C2262

Long divider

Confidence 1/5
Component profile — C2262 · layer 12 · positive tail

Logit-lens tokens · positive

#1"\r"#2"\"?>"#3"stretchr"#4" <<<<<<<<<<<<<<"#5"\">'.$"#6"writeFieldEnd"#7"WriteTagHelper"#8"+#+"#9"Lähteet"#10" nahilalakip"

R-lens tokens · positive

#1" quæ"#2" bonté"#3" cœurs"#4" peines"#5" eorum"#6" écrits"#7"</i>"#8" podjet"#9"   "#10" vérit"

Skewness

+0.06

Excess kurtosis

0.92rank #2110

Logcosh deviation

0.0091rank #2263

Top-score occurrences · positive

  1. "point"NameWeightColorSmellpoint (° C.)Si↕ +10.62 (#5) · ⚡ 1.6%
  2. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;↵;;↕ +10.50 (#4) · ⚡ 2.2%
  3. "----------------"------------------------------------------------------------------------------------------------------------------------------------------------↕ +10.31 (#6) · ⚡ 2.6%
  4. " *" * * * * * *↵ * ↕ +10.31 (#5) · ⚡ 2.8%
  5. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;↕ +10.31 (#5) · ⚡ 2.8%
  6. "////////////////"/////////////////////////////////////////////////////////////////////////////////////////////////↵↵:↕ +10.31 (#5) · ⚡ 2.3%
  7. "----------------"------------------------------------------------------------------------------------------------------------------------------------------------↕ +10.00 (#5) · ⚡ 2.6%
  8. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;↕ +9.94 (#5) · ⚡ 2.8%
  9. "\n\n"<bos>Description↵↵Fleshlight is proud↕ +9.56 (#32) · ⚡ 0.4%
  10. "\n\n"<bos>Description↵↵The one plugin you↕ +9.56 (#32) · ⚡ 0.4%
  11. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;↕ +9.56 (#6) · ⚡ 2.5%
  12. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;↕ +9.44 (#7) · ⚡ 2.3%
  13. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;↵;;;↕ +9.44 (#6) · ⚡ 2.2%
  14. " *" * * * * * //↵↵namespace Foam↕ +9.31 (#6) · ⚡ 2.4%
  15. " Orphan" Stakeholders Respect for Each Orphan Diseases Individual’s↕ +9.19 (#4) · ⚡ 1.4%
  16. "3"333333333↕ +9.06 (#7) · ⚡ 1.9%
  17. " *" * * * * * * * * *↕ +9.06 (#6) · ⚡ 2.3%
  18. " the" Each Other to at the ForefrontWe push↕ +9.00 (#2) · ⚡ 2.2%
  19. " *" * * * * * * */↵↵#↕ +8.94 (#5) · ⚡ 2.2%
  20. ";;;;";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;↕ +8.94 (#7) · ⚡ 2.2%
Gemma 2 2B · L13/C210

Toy/example code

Confidence 5/5
Component profile — C210 · layer 13 · negative tail

Logit-lens tokens · negative

#1"djangoproject"#2"MessageTagHelper"#3"цездатний"#4"argli"#5"ainville"#6"AndEndTag"#7"جوايز"#8" AssemblyProduct"#9" XYZ"#10"apimachinery"

R-lens tokens · negative

#1" stället"#2" fumée"#3" jouets"#4" inconvénients"#5" varandra"#6" colonnes"#7" mères"#8" normaux"#9" étoient"#10" démocr"

Skewness

-3.97

Excess kurtosis

45.84rank #422

Logcosh deviation

0.0947rank #211

Top-score occurrences · negative

  1. "Hello" [STRING] [Hello World, I would↕ -25.75 (#1) · ⚡ 22.5%
  2. " java"List;↵import java.util.Map↕ -25.62 (#1) · ⚡ 24.8%
  3. "topic", topic: ["topic_1", "↕ -25.00 (#1) · ⚡ 18.7%
  4. " ('"__slots__ = ('a', 'b↕ -23.12 (#1) · ⚡ 17.4%
  5. "\n\n"private String blog;↵↵ private String github↕ -23.12 (#1) · ⚡ 21.1%
  6. "a"tpUrl(), "a", Exchange.FILE↕ -23.00 (#1) · ⚡ 21.3%
  7. "var" public static $var;↵}↵↵↕ -22.75 (#1) · ⚡ 17.6%
  8. "baz" return bar.baz.operation()↵↕ -22.62 (#1) · ⚡ 16.4%
  9. "foo"<unq:foo xmlns:unq↕ -22.25 (#1) · ⚡ 15.0%
  10. "helloworld" []={"cout<<"helloworld " ;} ↕ -22.12 (#1) · ⚡ 15.8%
  11. "value" = {↵ value: 0;↕ -22.12 (#1) · ⚡ 18.6%
  12. "\n"Text(filename);↵ sw.WriteLine↕ -22.00 (#1) · ⚡ 18.8%
  13. "message" @Value("${message:Hello default}")↕ -21.75 (#1) · ⚡ 18.4%
  14. "Hello"C1', 'Hello');↵$objPHPExcel↕ -21.75 (#1) · ⚡ 14.2%
  15. "This" <div>This is arabic!</div↕ -21.38 (#1) · ⚡ 17.0%
  16. "-").data('current-step')+ "↕ -21.12 (#1) · ⚡ 15.7%
  17. "hello" a message like "hello world" to it↕ -21.12 (#1) · ⚡ 18.6%
  18. "Hello" - the string "Hello" is represented as↕ -21.12 (#1) · ⚡ 13.6%
  19. "\n".ConfigurationProperties;↵import org.springframework↕ -20.88 (#1) · ⚡ 13.4%
  20. "Hello" a simple write('Hello world'); and then↕ -20.75 (#1) · ⚡ 15.4%
Gemma 2 2B · L13/C528

So

Confidence 5/5
Component profile — C528 · layer 13 · positive tail

Logit-lens tokens · positive

#1" So"#2" so"#3"So"#4" SO"#5"so"#6"SO"#7"Sooo"#8"tso"#9" Sio"#10" אז"

R-lens tokens · positive

#1" SO"#2" So"#3" Sox"#4"So"#5"SO"#6"Sooo"#7"tso"#8" sooo"#9" Soo"#10"pso"

Skewness

+3.43

Excess kurtosis

53.65rank #348

Logcosh deviation

0.0687rank #529

Top-score occurrences · positive

  1. " So" when originally written. So why does the Republican↕ +22.00 (#1) · ⚡ 23.4%
  2. "So" it personally.↵↵So, how do you↕ +21.62 (#1) · ⚡ 20.7%
  3. " So" your question interesting. So +1 from me↕ +21.50 (#1) · ⚡ 19.8%
  4. " So"t my own. So I was thrown into↕ +21.25 (#1) · ⚡ 18.3%
  5. " So" hear about it! So far, this seems↕ +21.00 (#1) · ⚡ 23.2%
  6. "So" matrix rows)? So far it's↕ +20.88 (#1) · ⚡ 19.9%
  7. " So" and Eighth Avenue. So he knows nothing.↕ +20.88 (#1) · ⚡ 18.3%
  8. " so" for greater understanding, so that people of color↕ +20.75 (#1) · ⚡ 19.9%
  9. "So" a tool.↵↵So in looking back at↕ +20.62 (#1) · ⚡ 20.0%
  10. "So" abrupt change.↵↵So the government should stop↕ +20.62 (#1) · ⚡ 19.3%
  11. " So" us?”€ So he shot her.↕ +20.62 (#1) · ⚡ 16.7%
  12. "So" for it.↵↵So last decade↵↵The↕ +20.62 (#1) · ⚡ 18.8%
  13. "So" that way.↵↵So, anyway–lots↕ +20.62 (#1) · ⚡ 20.0%
  14. " So"’t we? So why not combine the↕ +20.50 (#1) · ⚡ 12.4%
  15. "So"↵↵Quote:↵↵So your old therapist is↕ +20.38 (#1) · ⚡ 18.8%
  16. " So" the new year. So in this case what↕ +20.38 (#1) · ⚡ 16.9%
  17. " So"led into fists. So he says, softly↕ +20.38 (#1) · ⚡ 18.8%
  18. "So" with Twilight.↵↵So, to start:↕ +20.25 (#1) · ⚡ 18.8%
  19. "So" LinkedIn profile.↵↵So far, it appears↕ +20.25 (#1) · ⚡ 20.8%
  20. "So" not careful!↵↵So you’re looking↕ +20.25 (#1) · ⚡ 15.9%
Gemma 2 2B · L14/C211

Second token position

Confidence 4/5
Component profile — C211 · layer 14 · positive tail

Logit-lens tokens · positive

#1" "#2"\n\n"#3" The"#4" A"#5" a"#6" the"#7"\n"#8" \""#9" I"#10"."

R-lens tokens · positive

#1" "#2"\n\n"#3" “"#4"\n"#5"<strong>"#6" "#7" The"#8" I"#9" New"#10" ‘"

Skewness

+4.76

Excess kurtosis

59.64rank #401

Logcosh deviation

0.1049rank #212

Top-score occurrences · positive

  1. " is"<bos>This is a walk on the↕ +23.88 (#2) · ⚡ 15.4%
  2. " have"<bos>We have all experienced that creepy↕ +23.25 (#1) · ⚡ 14.6%
  3. " have"<bos>We have studied the application of↕ +23.25 (#1) · ⚡ 14.6%
  4. " have"<bos>We have learned that our valued↕ +23.25 (#1) · ⚡ 14.5%
  5. " the"<bos>From the "child-like↕ +22.38 (#2) · ⚡ 10.4%
  6. " the"<bos>From the standpoint of comfort,↕ +22.38 (#2) · ⚡ 10.4%
  7. " the"<bos>From the mid-19↕ +22.38 (#2) · ⚡ 10.4%
  8. "2"<bos>1265 in Scotland↕ +22.12 (#1) · ⚡ 16.1%
  9. "2"<bos>123 HP OJPro↕ +22.12 (#1) · ⚡ 16.1%
  10. "9"<bos>1912 Müref↕ +22.00 (#1) · ⚡ 17.1%
  11. "9"<bos>1942 Iowa Pre↕ +22.00 (#1) · ⚡ 17.1%
  12. "2"<bos>428 F.2↕ +22.00 (#1) · ⚡ 18.8%
  13. "-"<bos>2-styrylchromones↕ +21.88 (#1) · ⚡ 12.7%
  14. "3"<bos>535 F.2↕ +21.75 (#1) · ⚡ 16.1%
  15. "7"<bos>176 page of collection↕ +21.75 (#1) · ⚡ 18.2%
  16. "0"<bos>10800 Zombies↕ +21.75 (#1) · ⚡ 18.0%
  17. "6"<bos>162 F.3↕ +21.62 (#1) · ⚡ 17.6%
  18. " to"<bos>Welcome to nose2↵↵nose↕ +21.50 (#1) · ⚡ 8.7%
  19. " are"<bos>We are excited that you will↕ +21.50 (#1) · ⚡ 14.1%
  20. " to"<bos>Welcome to the introduction to the↕ +21.50 (#1) · ⚡ 8.7%
Gemma 2 2B · L14/C1109

Data schema / field type definition

Confidence 5/5
Component profile — C1109 · layer 14 · positive tail

Logit-lens tokens · positive

#1"IntoConstraints"#2"ConstraintMaker"#3"GEBURTS"#4"usermodel"#5"imageshack"#6"IBarButtonItem"#7"ainville"#8"tences"#9" //$"#10")‏"

R-lens tokens · positive

#1" modalités"#2" négociations"#3" travailleurs"#4" investissements"#5" dépenses"#6" peines"#7" déclarations"#8" valeur"#9" prochaines"#10" renseignements"

Skewness

+1.94

Excess kurtosis

24.06rank #1011

Logcosh deviation

0.0506rank #1110

Top-score occurrences · positive

  1. "',".Column('amount', NumpyType)↕ +28.38 (#1) · ⚡ 35.5%
  2. ":" {↵ denom: PropTypes.object.↕ +28.12 (#1) · ⚡ 25.6%
  3. "," field :id, :id↵ ↕ +26.88 (#1) · ⚡ 28.3%
  4. " {"↵↵ visible: {↵ type:↕ +25.00 (#1) · ⚡ 23.9%
  5. " int" TABLE Member (Id int, Points int);↕ +24.62 (#1) · ⚡ 14.1%
  6. ":" {↵ type: Array,↵ ↕ +24.25 (#1) · ⚡ 22.9%
  7. " int"Feed_ID` int(11)↕ +22.25 (#1) · ⚡ 13.9%
  8. " ="'↵↵ name = fields.Char(↕ +22.12 (#1) · ⚡ 21.6%
  9. "int"phone_number bigint,↵ email↕ +21.88 (#1) · ⚡ 14.7%
  10. " " "Ordered",↵ "dataType": "↕ +21.88 (#1) · ⚡ 22.9%
  11. "]"Action] [int] NOT NULL, ↕ +21.62 (#1) · ⚡ 24.1%
  12. "type": {↵ type: Function,↵↕ +21.62 (#1) · ⚡ 12.3%
  13. "," arg :id, :id↵↵ ↕ +21.25 (#1) · ⚡ 20.3%
  14. " \"" "type": "string",↵ ↕ +21.25 (#1) · ⚡ 23.3%
  15. "," field :users, list_of(:↕ +20.75 (#1) · ⚡ 19.3%
  16. "fiscal"nro_nota_fiscal integer,↵ ↕ +20.50 (#1) · ⚡ 13.3%
  17. "1"` int(11) NOT NULL,↕ +20.50 (#1) · ⚡ 17.3%
  18. "\""tip" ("id" integer PRIMARY KEY AUTO↕ +20.38 (#1) · ⚡ 17.1%
  19. " NULL"" bigint NOT NULL DEFAULT (0),↕ +20.38 (#1) · ⚡ 13.3%
  20. "string"email" type="string" not-null↕ +20.12 (#1) · ⚡ 18.2%
Gemma 2 2B · L14/C1846

?

Confidence 0/5
Component profile — C1846 · layer 14 · negative tail

Logit-lens tokens · negative

#1" ſtate"#2" Majefty"#3"endphp"#4" ſeveral"#5"UrlResolution"#6" pleaſure"#7" beſt"#8" preſent"#9" ſte"#10" houſe"

R-lens tokens · negative

#1" practical"#2" getItemCount"#3"endphp"#4" roadside"#5"########."#6" المعيارى"#7"UIApplication"#8" physically"#9" itinéraires"#10" transportation"

Skewness

-0.54

Excess kurtosis

3.13rank #2001

Logcosh deviation

0.0262rank #1847

Top-score occurrences · negative

  1. " and" CYP1A genes and led to changes in↕ -15.12 (#1) · ⚡ 10.4%
  2. "," Orme, III, Larry W. Penn↕ -12.75 (#1) · ⚡ 7.1%
  3. "\t" = &sch↵ }↵ if↕ -12.25 (#3) · ⚡ 5.4%
  4. "\n\n"ries↵ )↵↵ acr_cred↕ -11.75 (#3) · ⚡ 4.2%
  5. " the"Count is larger than the actual RowCount,↕ -10.81 (#2) · ⚡ 5.0%
  6. " and" HIV and with attitude and behavior changes that would↕ -10.69 (#3) · ⚡ 5.4%
  7. " and" professionals enrich their theories and practice to advance their↕ -10.69 (#2) · ⚡ 6.3%
  8. "." keras.layers.AvgPool2D↕ -10.44 (#1) · ⚡ 5.5%
  9. "_" try { m_streamwriter.Dispose↕ -10.31 (#2) · ⚡ 4.5%
  10. " in" Mexico as they have in the United States but↕ -10.12 (#1) · ⚡ 3.8%
  11. " as" and why as well as ideas on how to↕ -10.06 (#1) · ⚡ 3.1%
  12. " a" a hex string from a Buffer.↵ */↕ -10.00 (#1) · ⚡ 2.5%
  13. "\n\n"var;↵}↵↵class X↵{↕ -10.00 (#1) · ⚡ 4.5%
  14. "\n"=etaCha,↵ final etaDis↕ -9.94 (#2) · ⚡ 3.8%
  15. "." of salinity is possible. In spite of this↕ -9.75 (#2) · ⚡ 4.6%
  16. "." susceptible to MDR mechanisms. The median IC9↕ -9.56 (#2) · ⚡ 4.6%
  17. "import".ElementTree↵import pandas as pd↵↵↕ -9.56 (#5) · ⚡ 3.0%
  18. "," of more complex skills, and actual golf putting↕ -9.44 (#2) · ⚡ 3.9%
  19. " REG"↵↵ datatype = REG_SZ; datas↕ -9.38 (#1) · ⚡ 2.5%
  20. " m"↵ try { m_streamwriter.↕ -9.38 (#2) · ⚡ 3.3%
Gemma 2 2B · L14/C1937

Conditional compilation in C/C++ and other languages

Confidence 3/5
Component profile — C1937 · layer 14 · negative tail

Logit-lens tokens · negative

#1"tagHelperRunner"#2"enumi"#3"IUrlHelper"#4"Diweddarwch"#5"AnchorStyles"#6"SequentialGroup"#7" חיצוניים"#8"دانشنامهٔ"#9" مشين"#10" ***!"

R-lens tokens · negative

#1" civilisation"#2"Fibre"#3"<code>"#4" fertil"#5"organised"#6"Colour"#7" Honourable"#8"colours"#9" privatisation"#10"organisation"

Skewness

-0.40

Excess kurtosis

5.83rank #1849

Logcosh deviation

0.0222rank #1938

Top-score occurrences · negative

  1. "]"_half_last]↵↵[box type↕ -18.62 (#1) · ⚡ 8.9%
  2. "\n"_TMH)↵↵#include "D↕ -18.50 (#1) · ⚡ 12.2%
  3. "\n" _WIN32↵#include "win↕ -17.62 (#1) · ⚡ 12.7%
  4. "\n"ING_UBL)↵ #include "../../↕ -17.50 (#1) · ⚡ 10.9%
  5. "\n"ifdef DEBUG_PCI↵#define PCI_↕ -17.12 (#1) · ⚡ 9.8%
  6. "\n"_HOST_H↵#define PCI_↕ -16.88 (#2) · ⚡ 7.5%
  7. "\n" __OBJC__↵#import <UIKit↕ -16.88 (#1) · ⚡ 9.8%
  8. "\n")↵#else↵#define PCI_↕ -16.75 (#1) · ⚡ 8.1%
  9. "\n"NO_REFLECTION↵#include "G↕ -16.38 (#1) · ⚡ 8.0%
  10. "\n"teensy41↵ static DirFat↕ -16.38 (#2) · ⚡ 10.7%
  11. "\n" HANDLE↵#else↵public typealias Thread↕ -16.38 (#1) · ⚡ 8.6%
  12. "\n"_TEENSY↵#define HostOS↕ -16.38 (#1) · ⚡ 9.7%
  13. "\n"ifndef FOUNDATION_EXPORT↵#if defined(__↕ -16.25 (#1) · ⚡ 8.3%
  14. "\n"PROCESSING_MODE)↵# include <↕ -16.25 (#1) · ⚡ 10.0%
  15. "\n"USER_ONLY)↵#include "soft↕ -16.25 (#1) · ⚡ 10.6%
  16. "\n"#region MyCode↵ public class Some↕ -16.12 (#1) · ⚡ 7.2%
  17. "\n"()↵#else↵ #define RAD↕ -16.00 (#1) · ⚡ 8.4%
  18. "\"?>")="x86"?> ↵ <↕ -15.69 (#2) · ⚡ 6.8%
  19. "\n"MSC_VER)↵# pragma once↕ -15.62 (#2) · ⚡ 9.2%
  20. "\n";↵#else↵typedef unsigned long long↕ -15.25 (#1) · ⚡ 7.7%
Gemma 2 2B · L15/C1900

Present

Confidence 5/5
Component profile — C1900 · layer 15 · negative tail

Logit-lens tokens · negative

#1" presented"#2" presenting"#3" Presented"#4" presents"#5" presentation"#6"Presented"#7" presenta"#8" Presenting"#9" Presenta"#10" presentations"

R-lens tokens · negative

#1" presenting"#2" Presenting"#3" presentation"#4" presented"#5" presentations"#6"Presenting"#7" Presented"#8" Presentation"#9"Presented"#10" presenters"

Skewness

-0.89

Excess kurtosis

8.93rank #1888

Logcosh deviation

0.0320rank #1901

Top-score occurrences · negative

  1. " presents"↵If the attacker presents a certificate, but↕ -21.38 (#1) · ⚡ 13.5%
  2. " present" conventional methods, will present the results before the↕ -20.50 (#1) · ⚡ 13.2%
  3. " presented" in which Galland presented his view on what↕ -19.25 (#1) · ⚡ 14.0%
  4. " presents" that when pressed, presents you with a random↕ -19.25 (#1) · ⚡ 12.1%
  5. " presents" NED this month, presents a compelling case that↕ -18.75 (#1) · ⚡ 12.7%
  6. " presented" Trumai, has presented varied inventories of these↕ -18.62 (#1) · ⚡ 14.4%
  7. " presents" sports and more and presents them in ways that↕ -18.12 (#1) · ⚡ 11.1%
  8. " presented" loan application will be presented to the Loan Review↕ -17.62 (#1) · ⚡ 10.1%
  9. " presents" this tuning fraternity and presents the↵Widestar↕ -17.50 (#1) · ⚡ 10.1%
  10. " presented"'s and are presented in a number of↕ -16.50 (#1) · ⚡ 10.9%
  11. " presented"16 years, presented with clinical stages I↕ -16.38 (#1) · ⚡ 9.2%
  12. " present", so as to present a smaller hoop within↕ -16.38 (#1) · ⚡ 10.8%
  13. " present" particular taxon, and present quantitative data regarding relative↕ -16.38 (#1) · ⚡ 11.5%
  14. " presented"-old male child presented with a right-↕ -16.12 (#2) · ⚡ 8.1%
  15. " presents" on it, it presents you with a clipboard↕ -16.00 (#1) · ⚡ 9.1%
  16. " presenting" sue for peace, presenting it as 'the↕ -15.94 (#1) · ⚡ 8.3%
  17. " presented" of the group yesterday presented a portrait of O↕ -15.88 (#1) · ⚡ 7.6%
  18. " presented" user will then be presented with an image selector↕ -15.75 (#1) · ⚡ 8.3%
  19. " presented" previously, he had presented with flu-like↕ -15.69 (#2) · ⚡ 8.3%
  20. " present" 10 and present their document to Mrs↕ -15.44 (#1) · ⚡ 7.4%
Gemma 2 2B · L18/C1304

Nuclear medicine imaging

Confidence 5/5
Component profile — C1304 · layer 18 · positive tail

Logit-lens tokens · positive

#1" radioactivity"#2" radioactive"#3"RADIO"#4" radio"#5" Radioactive"#6"radio"#7" nuclear"#8" RADIO"#9" Nuclear"#10"Nuclear"

R-lens tokens · positive

#1" radioactive"#2" radioactivity"#3" radionuc"#4" nuclear"#5" Radioactive"#6" radio"#7" Nuclear"#8"radio"#9"Nuclear"#10" Radio"

Skewness

+3.27

Excess kurtosis

60.50rank #839

Logcosh deviation

0.0614rank #1305

Top-score occurrences · positive

  1. " "1 keV and at 245 keV↕ +32.00 (#1) · ⚡ 32.8%
  2. "1" of In-111, which has↕ +31.62 (#1) · ⚡ 27.3%
  3. "O"Tc-HMPAO tumor scintigraphy↕ +30.12 (#1) · ⚡ 32.0%
  4. "2"guanidine ((123)I-↕ +29.75 (#1) · ⚡ 28.1%
  5. " of"5 keV; complexes of Ga-67↕ +29.75 (#1) · ⚡ 34.8%
  6. " include" the radiopharmaceutical may include a radioisotope attached↕ +29.38 (#1) · ⚡ 36.5%
  7. "2" for iodine-123-metaiod↕ +29.38 (#1) · ⚡ 36.3%
  8. "Tc" 99mTc-hexamethyl↕ +29.00 (#1) · ⚡ 26.6%
  9. " nuclear" in other types of nuclear imaging, nuclear spectroscopy↕ +28.50 (#1) · ⚡ 22.9%
  10. " in" or a general increase in radio accumulation throughout the↕ +28.25 (#1) · ⚡ 32.4%
  11. " radio" body in which the radiopharmaceutical is concentrated.↕ +28.12 (#1) · ⚡ 23.5%
  12. "idine"iodobenzylguanidine imaging protocol in heart↕ +28.12 (#1) · ⚡ 26.0%
  13. " "93 keV, 184 keV↕ +27.75 (#1) · ⚡ 26.6%
  14. "9" using Tc-99m MAG3 differ↕ +27.50 (#1) · ⚡ 27.5%
  15. "-"99mTc-HMPAO tumor↕ +27.50 (#1) · ⚡ 34.2%
  16. "3" iodine-123-metaiodoben↕ +27.50 (#1) · ⚡ 29.5%
  17. ",", gallium scans, indium white blood↕ +27.50 (#1) · ⚡ 35.5%
  18. "9"2 mCi 99Tc-H↕ +27.38 (#1) · ⚡ 28.9%
  19. " abnormally" Most previous reports involve abnormally increased accumulation of radio↕ +27.25 (#1) · ⚡ 26.6%
  20. "uc"].↵Radionuclide renal study is↕ +27.25 (#1) · ⚡ 27.7%
Gemma 2 2B · L18/C1345

Sentence-final additive/concessive adverb pattern, anyway

Confidence 5/5
Component profile — C1345 · layer 18 · positive tail

Logit-lens tokens · positive

#1"."#2","#3"!"#4"帖最后由"#5";"#6" grà"#7" irré"#8" because"#9" gambe"#10" réessayer"

R-lens tokens · positive

#1"."#2","#3"!"#4";"#5" because"#6"!."#7" –"#8" but"#9"!,"#10" while"

Skewness

+2.27

Excess kurtosis

22.12rank #1710

Logcosh deviation

0.0598rank #1346

Top-score occurrences · positive

  1. " too"'s Travels," too, but I don↕ +19.75 (#1) · ⚡ 14.4%
  2. " too" fun with it, too. It uses coding↕ +19.12 (#1) · ⚡ 14.5%
  3. " too" Southeast Alaska coastal waters too. Following the migrating↕ +19.00 (#1) · ⚡ 15.0%
  4. " though" won't have though, so you may↕ +18.62 (#1) · ⚡ 15.8%
  5. " too" for incumbents, too. Ping An,↕ +18.38 (#1) · ⚡ 14.5%
  6. " too"ying is good, too. Gay loudmouth↕ +18.12 (#1) · ⚡ 16.1%
  7. " anyway"s the story, anyway.↵↵Step ↕ +18.00 (#1) · ⚡ 14.6%
  8. " anyway" stories were completed, anyway). At the same↕ +18.00 (#1) · ⚡ 19.3%
  9. " though" have an electrical charge though, which means they↕ +18.00 (#1) · ⚡ 14.6%
  10. " course" More football, of course.↵↵Matchday↕ +17.88 (#1) · ⚡ 14.1%
  11. " though"grind thinks you are though, and for this↕ +17.88 (#1) · ⚡ 14.8%
  12. " too" usually a singer, too, and often provides↕ +17.88 (#1) · ⚡ 12.9%
  13. " though" part of the calculation though - we need to↕ +17.50 (#1) · ⚡ 14.1%
  14. " anyway"-rated brains, anyway.↵↵We'↕ +17.38 (#1) · ⚡ 16.1%
  15. " too" in the aesthetic department too. So far,↕ +17.38 (#1) · ⚡ 15.8%
  16. " too" for less secure leaders too. Any moderate competence↕ +17.12 (#1) · ⚡ 14.1%
  17. " either" way to do it either.↵↵The GOP↕ +17.12 (#1) · ⚡ 16.1%
  18. " too" the state level, too.↵↵ROUSE↕ +17.00 (#1) · ⚡ 15.1%
  19. " either"t exactly instill confidence either. Many hawks on↕ +17.00 (#1) · ⚡ 15.4%
  20. " anyway" the Legends Canon, anyway) as the Ald↕ +16.88 (#1) · ⚡ 17.1%
Gemma 2 2B · L20/C86

C

Confidence 5/5
Component profile — C86 · layer 20 · positive tail

Logit-lens tokens · positive

#1" getC"#2" albicans"#3" cérami"#4" ALC"#5" Merv"#6" Cæsar"#7" elegans"#8" Mep"#9" normaux"#10"setC"

R-lens tokens · positive

#1" prêtres"#2" cérami"#3" stället"#4" guerres"#5" prochaines"#6" supérieurs"#7" producteurs"#8" normaux"#9" complètes"#10" colorés"

Skewness

+12.08

Excess kurtosis

302.83rank #70

Logcosh deviation

0.1678rank #87

Top-score occurrences · positive

  1. "C"fasteo↵Cataract surgery is↕ +45.75 (#1) · ⚡ 52.3%
  2. "C" consumer sentiment.↵↵Capping off the bad↕ +43.00 (#1) · ⚡ 49.8%
  3. " C" or How Reddit Will Censor Controversial Content↕ +40.50 (#1) · ⚡ 49.0%
  4. " C"/**↵ * Caches requests for Media↕ +40.00 (#1) · ⚡ 49.8%
  5. " C":↵↵What is C#.net meaning in↕ +40.00 (#1) · ⚡ 51.6%
  6. "C" desserts↵Category:Cieszyn↵Category↕ +39.50 (#1) · ⚡ 43.4%
  7. "c" them that their “cures” are not↕ +39.00 (#1) · ⚡ 48.0%
  8. " C" this week’s CROM’s Corner↕ +38.75 (#1) · ⚡ 50.4%
  9. " c"<bos>Can cN0 and p↕ +38.25 (#1) · ⚡ 52.7%
  10. " C" journalism.↵↵As CJR reported in March↕ +38.00 (#1) · ⚡ 35.7%
  11. "C"8↵↵Experience:Claddagh Wedding↕ +38.00 (#1) · ⚡ 44.3%
  12. "C"next build:↵↵C:\path\to↕ +37.25 (#1) · ⚡ 45.1%
  13. "c".55↵↵↵↵cdebconf↵↵↵↵db↕ +37.25 (#1) · ⚡ 49.8%
  14. " c" growth so why not cialis 5mg↕ +36.50 (#1) · ⚡ 34.6%
  15. " C"m working on a C# webforms application↕ +36.50 (#1) · ⚡ 44.7%
  16. " C" when it crashes, C++ is returning a↕ +36.50 (#1) · ⚡ 44.1%
  17. " c" to a server using cURL without any headers↕ +36.25 (#1) · ⚡ 43.9%
  18. "C"85↵ Clem Frobisher↕ +36.25 (#1) · ⚡ 44.1%
  19. " C"↵the source is C# .Net ↕ +36.25 (#1) · ⚡ 44.5%
  20. " C" Role Modeling" and CQL are actually a↕ +35.75 (#1) · ⚡ 39.5%
Gemma 2 2B · L20/C736

Banking / lending

Confidence 5/5
Component profile — C736 · layer 20 · positive tail

Logit-lens tokens · positive

#1" bank"#2" banks"#3" banking"#4" Bank"#5" loan"#6" lending"#7" loans"#8" bankers"#9" lenders"#10"bank"

R-lens tokens · positive

#1" bank"#2" banks"#3" banking"#4" Bank"#5" Banking"#6" bankers"#7" loans"#8" loan"#9"Bank"#10"bank"

Skewness

+3.52

Excess kurtosis

43.39rank #1094

Logcosh deviation

0.0831rank #737

Top-score occurrences · positive

  1. " generous" the two received overly generous terms and conditions on↕ +25.62 (#1) · ⚡ 22.2%
  2. " for" institutions charge each other for extremely short-term↕ +23.50 (#1) · ⚡ 14.3%
  3. " to" be no further reforms to increase access to credit↕ +22.62 (#1) · ⚡ 16.4%
  4. " at" challenges are particularly acute at smaller banks and credit↕ +22.50 (#1) · ⚡ 19.1%
  5. " retail", private bank, retail bank, family office↕ +22.50 (#1) · ⚡ 14.9%
  6. " deposit", head of consumer deposit products at TD Bank↕ +22.38 (#1) · ⚡ 14.7%
  7. "," for the investment management, private bank, retail↕ +22.00 (#1) · ⚡ 17.5%
  8. " and" acute at smaller banks and credit unions, which↕ +21.88 (#1) · ⚡ 14.3%
  9. " because" puny these days because rates are being held↕ +21.62 (#1) · ⚡ 13.1%
  10. " for" than they can use for lending, and the↕ +21.50 (#1) · ⚡ 17.5%
  11. " of" as Assistant Vice President of Corporate Banking, Mr↕ +21.50 (#1) · ⚡ 16.2%
  12. " bank" to look at their bank accounts, see interest↕ +21.50 (#1) · ⚡ 11.7%
  13. " to" banks are less willing to lend to individuals who↕ +21.25 (#1) · ⚡ 16.1%
  14. " guarantee" bankrupt and refused to guarantee bank deposits? Where↕ +21.25 (#1) · ⚡ 14.5%
  15. " to" including those without access to traditional financial services.”↕ +21.12 (#1) · ⚡ 13.8%
  16. " and" deployment for international retail and private bank. The↕ +21.00 (#1) · ⚡ 12.7%
  17. " higher"-only banks offer higher interest rates on their↕ +20.88 (#1) · ⚡ 14.6%
  18. "4" PersonalLoans↵↵#4 iCashLoans↵↵↕ +20.88 (#1) · ⚡ 9.5%
  19. " Interest" with a free High Interest Checking Account from Bank↕ +20.75 (#1) · ⚡ 11.1%
  20. " for", if the demand for loans increases, and↕ +20.38 (#1) · ⚡ 17.5%
Gemma 2 2B · L20/C949

Decorative aesthetics

Confidence 5/5
Component profile — C949 · layer 20 · negative tail

Logit-lens tokens · negative

#1" gorgeous"#2" beautiful"#3" lovely"#4" inspiration"#5" beautifully"#6" design"#7" adorable"#8" inspired"#9" amazing"#10" fabulous"

R-lens tokens · negative

#1" gorgeous"#2" beautiful"#3" adorable"#4" lovely"#5" inspiration"#6" decorating"#7" beautifully"#8" decorative"#9" cute"#10" décor"

Skewness

-2.87

Excess kurtosis

32.85rank #1328

Logcosh deviation

0.0739rank #950

Top-score occurrences · negative

  1. " those" the ribbon treatment and those gorgeous flowers!!! Beautiful↕ -25.62 (#1) · ⚡ 22.7%
  2. " the" love that image and the colurs you'↕ -25.50 (#1) · ⚡ 19.4%
  3. " to" from florals, to wood textures, murals↕ -23.00 (#1) · ⚡ 16.5%
  4. " and" of glittered lace and a few pearls.↕ -22.88 (#1) · ⚡ 16.0%
  5. " a" Up...just added a touch of glittered↕ -22.62 (#1) · ⚡ 17.5%
  6. " this".↵↵I found this monogram that Centsational↕ -22.38 (#1) · ⚡ 17.9%
  7. " added"in Up...just added a touch of glit↕ -21.50 (#1) · ⚡ 16.2%
  8. " is" in wallpaper and fabrics is definitely Thibaut.↕ -21.50 (#1) · ⚡ 11.9%
  9. " Elements" S.U. Elements of Style stamp set↕ -21.25 (#1) · ⚡ 16.9%
  10. " the" I'm digging the above inspiration pic.↕ -21.00 (#1) · ⚡ 17.2%
  11. " small" 2 of the small butterflies from the S↕ -20.88 (#1) · ⚡ 16.2%
  12. "." an S.U. script background a onto↕ -20.75 (#1) · ⚡ 14.7%
  13. "tered" a touch of glittered lace and a few↕ -20.50 (#1) · ⚡ 13.3%
  14. " with" wine color and adorned with a gold brass buckle↕ -20.50 (#1) · ⚡ 12.2%
  15. " inspired" Anderson↵↵The garden inspired wallpaper in this dining↕ -20.38 (#1) · ⚡ 17.0%
  16. " this" you how to create this 2 in ↕ -20.38 (#1) · ⚡ 18.3%
  17. " its"↵↵This wallpaper with its bold pattern adds personality↕ -20.12 (#1) · ⚡ 17.0%
  18. "." used S.U. designer paper which matches↕ -20.12 (#1) · ⚡ 13.1%
  19. " the" stash but I know the yellow paper is from↕ -20.12 (#1) · ⚡ 17.8%
  20. " the". And we love the effect!↵↵Related↕ -20.12 (#1) · ⚡ 16.3%
Gemma 2 2B · L20/C1154

1960s–1970s American culture/history

Confidence 5/5
Component profile — C1154 · layer 20 · positive tail

Logit-lens tokens · positive

#1" Vietnam"#2" hippie"#3" seventies"#4" sixties"#5" disco"#6"Vietnam"#7" Nixon"#8" Viet"#9" Beatles"#10"]++;"

R-lens tokens · positive

#1" Vietnam"#2" hippie"#3"Vietnam"#4" Nixon"#5" Viet"#6" sixties"#7" seventies"#8" Beatles"#9" disco"#10" VIET"

Skewness

+2.77

Excess kurtosis

41.62rank #1127

Logcosh deviation

0.0661rank #1155

Top-score occurrences · positive

  1. " President"76 by then President Gerald Ford, and↕ +30.50 (#1) · ⚡ 23.8%
  2. " The" at the height of The Summer of Love.↕ +28.88 (#1) · ⚡ 22.9%
  3. " against" started for the Yankees against the Reds, whose↕ +28.62 (#1) · ⚡ 23.7%
  4. "The") - arranger↵The Carnival - The Carnival↕ +27.50 (#1) · ⚡ 19.8%
  5. " on" made a guest appearance on The Dating Game in↕ +27.00 (#1) · ⚡ 20.8%
  6. " when"60s, when the mere existence of↕ +26.50 (#1) · ⚡ 27.7%
  7. " Plymouth"1971 Plymouth Hemi 'C↕ +26.50 (#1) · ⚡ 21.0%
  8. "\n"972)↵The Crazy World of↕ +25.88 (#1) · ⚡ 22.2%
  9. " the" its debut appearance at the Monterey Pop Festival,↕ +25.62 (#1) · ⚡ 16.6%
  10. " The"966, The Beatles had taken refuge↕ +25.25 (#1) · ⚡ 19.1%
  11. "\n"4) - arranger↵Tom Waits - The↕ +24.50 (#1) · ⚡ 21.4%
  12. " after"65, shortly after Lyndon Johnson was sworn↕ +24.12 (#1) · ⚡ 15.1%
  13. "9"yled in 1970 by John↕ +24.00 (#1) · ⚡ 13.5%
  14. "," movies including Blow Up, Salome and Cole↕ +23.75 (#1) · ⚡ 17.5%
  15. " and" time, anyway), and Network, which earned↕ +23.75 (#1) · ⚡ 11.8%
  16. " The", from bands like The Beach Boys, started↕ +23.62 (#1) · ⚡ 14.5%
  17. ",". Another English band, The Who, inspired↕ +23.62 (#1) · ⚡ 15.8%
  18. " player" Austin (Universal contract player Lee Majors), who↕ +23.50 (#1) · ⚡ 14.2%
  19. " for" the sentimental career nod for Art Carney in Harry↕ +23.50 (#1) · ⚡ 13.4%
  20. " including" in several cult movies including Blow Up, Sal↕ +23.25 (#1) · ⚡ 16.5%
Gemma 2 2B · L21/C1514

){

Confidence 5/5
Component profile — C1514 · layer 21 · negative tail

Logit-lens tokens · negative

#1" nahilalakip"#2"ⓧ"#3" feroit"#4" purpoſe"#5" beſt"#6" reaſon"#7" auffi"#8" Waray"#9" leaſt"#10" متعلقه"

R-lens tokens · negative

#1"{"#2"}{"#3" {"#4":{"#5"{."#6"}."#7"({"#8"?}"#9"()}"#10"[{"

Skewness

-2.85

Excess kurtosis

58.89rank #837

Logcosh deviation

0.0543rank #1515

Top-score occurrences · negative

  1. "){"1](#F1){ref-type="↕ -35.50 (#1) · ⚡ 27.9%
  2. "){"a](#fig1){ref-type="↕ -35.25 (#1) · ⚡ 30.9%
  3. "){"1](#F1){ref-type="↕ -34.50 (#1) · ⚡ 26.4%
  4. "){"1](#F1){ref-type="↕ -34.25 (#1) · ⚡ 30.9%
  5. "){"1](#F1){ref-type="↕ -34.00 (#1) · ⚡ 28.1%
  6. "){"1](#fig1){ref-type="↕ -33.75 (#1) · ⚡ 27.0%
  7. "){"1](#fig1){ref-type="↕ -33.50 (#1) · ⚡ 27.0%
  8. "){"1](#Fig1){ref-type="↕ -33.50 (#1) · ⚡ 25.2%
  9. "){"1](#T1){ref-type="↕ -33.25 (#1) · ⚡ 29.3%
  10. "){"1](#F1){ref-type="↕ -33.25 (#1) · ⚡ 27.3%
  11. "){"0005){ref-type="↕ -33.25 (#1) · ⚡ 25.2%
  12. "){"1](#F1){ref-type="↕ -33.25 (#1) · ⚡ 27.0%
  13. "){"](#MOESM1){ref-type="↕ -32.75 (#1) · ⚡ 26.4%
  14. "){"21-1){ref-type="↕ -32.50 (#1) · ⚡ 26.0%
  15. "){"-189){ref-type="↕ -32.25 (#1) · ⚡ 28.7%
  16. "){"-102){ref-type="↕ -32.00 (#1) · ⚡ 28.9%
  17. "){",](#Fig1){ref-type="↕ -32.00 (#1) · ⚡ 23.5%
  18. "){"-028){ref-type="↕ -31.25 (#1) · ⚡ 28.3%
  19. "){"0001){ref-type="↕ -30.00 (#1) · ⚡ 26.0%
  20. "]{"---↵↵[****]{}↵↵ ----------------↕ -29.75 (#1) · ⚡ 19.0%
Gemma 2 2B · L22/C618

After

Confidence 5/5
Component profile — C618 · layer 22 · positive tail

Logit-lens tokens · positive

#1" the"#2" they"#3" completing"#4" completion"#5" being"#6" finishing"#7" receiving"#8" initial"#9" hearing"#10" their"

R-lens tokens · positive

#1" completion"#2" completing"#3" graduation"#4" finishing"#5" being"#6" receiving"#7" they"#8" the"#9" hearing"#10" initial"

Skewness

+4.80

Excess kurtosis

82.84rank #484

Logcosh deviation

0.0857rank #619

Top-score occurrences · positive

  1. " After"guinnings - After becoming a fan of↕ +29.75 (#1) · ⚡ 37.1%
  2. "After" as well?↵↵After all, it signifies↕ +28.50 (#1) · ⚡ 28.5%
  3. "after" consistency when making "after eight" chocolate↵↵↕ +27.75 (#1) · ⚡ 25.2%
  4. "After": dive in↵↵After MLK day and↕ +27.75 (#1) · ⚡ 39.5%
  5. "AFTER"<bos>Q:↵↵AFTER INSERT Trigger not firing↕ +27.12 (#1) · ⚡ 30.1%
  6. " After" in foreign countries. After receiving the necessary document↕ +27.12 (#1) · ⚡ 33.6%
  7. " After" from the Nazis. After the communists left Budapest↕ +27.00 (#1) · ⚡ 25.6%
  8. "After" The Princess Bride↵↵After an unfortunate turn of↕ +26.75 (#1) · ⚡ 32.8%
  9. "After"C)loser↵After scratches, a horse↕ +26.38 (#1) · ⚡ 26.8%
  10. " After" graduate school information. After all, it makes↕ +26.38 (#1) · ⚡ 29.3%
  11. "After" to A.↵↵After all if you prefer↕ +26.38 (#1) · ⚡ 29.3%
  12. " After" the white flash? After all the brain is↕ +26.25 (#1) · ⚡ 28.5%
  13. " After" from criminal interference. After all, fair elections↕ +26.00 (#1) · ⚡ 29.3%
  14. " After"747. After much head-scratch↕ +26.00 (#1) · ⚡ 28.3%
  15. " After" there any differences? After you factor out the↕ +25.88 (#1) · ⚡ 24.4%
  16. "After" be pleased.↵↵After six years, I↕ +25.75 (#1) · ⚡ 26.0%
  17. " After" in cardiology. After my years of experience↕ +25.50 (#1) · ⚡ 21.3%
  18. " after" open. This time after circling the block,↕ +25.50 (#1) · ⚡ 22.1%
  19. " After"Talking about myself. After the book launch and↕ +25.38 (#1) · ⚡ 30.3%
  20. "After" treasures.↵↵“After watching the movie,↕ +25.12 (#1) · ⚡ 27.1%
Gemma 2 2B · L22/C1380

Digit in numeric identifier

Confidence 4/5
Component profile — C1380 · layer 22 · positive tail

Logit-lens tokens · positive

#1"0"#2"4"#3"7"#4"8"#5"9"#6"6"#7"3"#8"5"#9"2"#10"1"

R-lens tokens · positive

#1"0"#2"8"#3"9"#4"7"#5"4"#6"6"#7"5"#8"3"#9" giudizio"#10"2"

Skewness

+2.11

Excess kurtosis

22.71rank #1558

Logcosh deviation

0.0553rank #1381

Top-score occurrences · positive

  1. "1"20701204](↕ +18.88 (#1) · ⚡ 16.4%
  2. "0"004207A1 and↕ +17.75 (#1) · ⚡ 12.4%
  3. "1"0002120-2↕ +17.62 (#1) · ⚡ 12.5%
  4. "0"77240121/]↕ +17.50 (#1) · ⚡ 11.9%
  5. "4"2346485 (International↕ +17.25 (#1) · ⚡ 11.0%
  6. "2"-0002120-↕ +17.12 (#1) · ⚡ 16.0%
  7. "3"-287318, for↕ +17.00 (#1) · ⚡ 19.3%
  8. "1"0003123-2↕ +17.00 (#1) · ⚡ 9.8%
  9. "6"-00360.↵United↕ +17.00 (#1) · ⚡ 16.5%
  10. "7"5506791/↵↵↕ +16.88 (#1) · ⚡ 12.0%
  11. "4"-207450, a↕ +16.75 (#1) · ⚡ 19.9%
  12. "4"0.514882 To↕ +16.75 (#1) · ⚡ 11.7%
  13. "0"-00302-CR↵↵↕ +16.75 (#1) · ⚡ 13.6%
  14. "5"0-015-082↕ +16.75 (#1) · ⚡ 14.2%
  15. "9"-253949 A.↕ +16.75 (#1) · ⚡ 16.6%
  16. "9" A113951. In↕ +16.75 (#1) · ⚡ 10.3%
  17. "7"0 067 949↕ +16.62 (#2) · ⚡ 10.6%
  18. "4"253949 A.↵↕ +16.62 (#1) · ⚡ 11.8%
  19. "9" RR00995-13↕ +16.12 (#1) · ⚡ 10.4%
  20. "1"475211519↕ +16.00 (#1) · ⚡ 13.4%
Gemma 2 2B · L22/C1458

Greek-letters in scientific notation

Confidence 5/5
Component profile — C1458 · layer 22 · positive tail

Logit-lens tokens · positive

#1" alpha"#2"TagMode"#3" DELTA"#4" gamma"#5" beta"#6" lambda"#7" betta"#8"AndEndTag"#9" Lambda"#10"GraphicsUnit"

R-lens tokens · positive

#1" alpha"#2" gamma"#3" beta"#4" α"#5" lambda"#6" DELTA"#7" Delta"#8"α"#9" alfa"#10" Gamma"

Skewness

+2.41

Excess kurtosis

37.61rank #1123

Logcosh deviation

0.0526rank #1459

Top-score occurrences · positive

  1. "1" level of IL-1β, TNF-↕ +26.88 (#1) · ⚡ 22.3%
  2. "3" glycogen synthase kinase 3 (GSK3↕ +26.50 (#1) · ⚡ 19.8%
  3. " factor") and transforming growth factor-beta(1↕ +25.88 (#1) · ⚡ 18.6%
  4. "-" of tumor necrosis factor-alpha on sexual activity↕ +25.88 (#1) · ⚡ 22.4%
  5. " receptor" proliferator-activated receptor gamma (PPAR↕ +25.25 (#1) · ⚡ 15.5%
  6. "-"inflammatory cytokines (TNF-α) (by↕ +25.25 (#1) · ⚡ 27.0%
  7. ".".S.C. §§ 84↕ +25.12 (#1) · ⚡ 19.7%
  8. "PAR"cological activation of PPAR gamma ameliorates vascular↕ +25.12 (#1) · ⚡ 18.7%
  9. " TGF"GF, VEGF and TGF. Upregulation↕ +24.88 (#1) · ⚡ 18.6%
  10. ".".S.C. § 37↕ +24.75 (#2) · ⚡ 20.1%
  11. "-", tumor necrosis factor-alpha [TNF-↕ +24.50 (#1) · ⚡ 22.8%
  12. ".".S.C. § 13↕ +24.38 (#2) · ⚡ 20.1%
  13. "3"3 (GSK3), and β-↕ +24.38 (#1) · ⚡ 21.8%
  14. "1" α, interleukin-1β, oncostatin↕ +24.00 (#1) · ⚡ 19.5%
  15. " IFN"0, TNF, IFN-γ and IL↕ +23.75 (#1) · ⚡ 16.8%
  16. "-" and transforming growth factor-beta (TGF-↕ +23.50 (#1) · ⚡ 18.5%
  17. "7" percutaneous 17beta-estradiol↕ +23.38 (#1) · ⚡ 14.4%
  18. "(" with exponential distributions Exp(1)↵$↕ +23.38 (#1) · ⚡ 14.3%
  19. "5" coli* strain DH5α cells (Tak↕ +23.38 (#1) · ⚡ 14.1%
  20. "-" factor alpha (TNF-alpha) and their↕ +23.25 (#1) · ⚡ 25.8%
Gemma 2 2B · L22/C1852

Am-, prefer capitalized, American

Confidence 5/5
Component profile — C1852 · layer 22 · negative tail

Logit-lens tokens · negative

#1"[::-"#2"melidir"#3"subpackage"#4"RSpec"#5"IsPostBack"#6" famiglie"#7" Samoa"#8"mybatisplus"#9" ammat"#10"/**/*."

R-lens tokens · negative

#1" переписи"#2"PasswordEncoder"#3"[::-"#4"subpackage"#5" Samoa"#6" makeStyles"#7"melidir"#8"ndia"#9"/*++"#10"stdbool"

Skewness

-2.12

Excess kurtosis

48.57rank #888

Logcosh deviation

0.0387rank #1853

Top-score occurrences · negative

  1. " Am" a pay day! Amateurs and the novice↕ -36.00 (#1) · ⚡ 19.4%
  2. " American"t attend the recent American Gear Manufacturers Association (↕ -35.25 (#1) · ⚡ 25.8%
  3. " Am" on this, see Amory Lovins talk↕ -34.50 (#1) · ⚡ 21.4%
  4. " American" instance, in The American Spectator, one Jeffrey↕ -31.75 (#1) · ⚡ 23.5%
  5. " American"Last year, the American Academy of Pediatrics revised↕ -31.50 (#1) · ⚡ 19.1%
  6. "Am" and Damascus.↵↵Amnesty has launched the↕ -31.25 (#1) · ⚡ 17.2%
  7. " American"500 from American Express gave the Phoenix↕ -31.12 (#1) · ⚡ 23.4%
  8. " American" suffocation. The American Academy of Pediatrics also↕ -30.88 (#1) · ⚡ 17.5%
  9. " American"'Journal of the American College of Cardiology,↕ -30.50 (#1) · ⚡ 16.6%
  10. " American"↵↵According to the American Cancer Society, prostate↕ -30.50 (#1) · ⚡ 19.3%
  11. " American" this goal,” the American Heart Association CEO Nancy↕ -30.38 (#1) · ⚡ 18.8%
  12. "American".↵↵Chevron↵↵American Home Products↵↵"↕ -30.25 (#1) · ⚡ 27.9%
  13. " American" Committee, said the American public is growing weary↕ -30.12 (#1) · ⚡ 22.0%
  14. " Am"ns et al., Am. J. Pathol↕ -29.88 (#1) · ⚡ 20.0%
  15. " American" Utah chapter of the American Society of Addiction Medicine↕ -29.62 (#1) · ⚡ 18.5%
  16. " American" and presentations from the American Diabetes Association Scientific Sessions↕ -28.88 (#1) · ⚡ 17.1%
  17. " American"'Journal of the American College of Cardiology,↕ -28.88 (#1) · ⚡ 15.4%
  18. "Am"758 (Amur catfish)↵↕ -28.75 (#1) · ⚡ 16.1%
  19. "American"-Appellee, andAmerican Telephone& Telegraph Company↕ -28.50 (#1) · ⚡ 19.9%
  20. "American" Raw” and “American Pickers” and↕ -28.38 (#1) · ⚡ 16.6%
Gemma 2 2B · L24/C507

7

Confidence 5/5
Component profile — C507 · layer 24 · positive tail

Logit-lens tokens · positive

#1"7"#2"8"#3"5"#4"9"#5"0"#6"6"#7"4"#8"2"#9"3"#10"1"

R-lens tokens · positive

#1"7"#2"5"#3"8"#4"9"#5"6"#6"0"#7"4"#8"3"#9"2"#10"1"

Skewness

+4.02

Excess kurtosis

64.19rank #486

Logcosh deviation

0.0819rank #508

Top-score occurrences · positive

  1. "7" until 1973.↕ +24.88 (#1) · ⚡ 38.1%
  2. "7" in 1973.↵↵Expansion↕ +24.62 (#1) · ⚡ 37.5%
  3. "7"March 1971 ↵D↕ +24.62 (#1) · ⚡ 39.1%
  4. "7") 1976, p.↕ +24.38 (#1) · ⚡ 35.0%
  5. "7"9%). Sixteen (73%) patients underwent↕ +24.25 (#1) · ⚡ 12.8%
  6. "7" the 1971 Franco Rubar↕ +24.12 (#1) · ⚡ 12.3%
  7. "7" o?↵-72*o*↕ +24.00 (#1) · ⚡ 13.5%
  8. "7"original_height":720,"content↕ +24.00 (#1) · ⚡ 21.2%
  9. "7" February 1971):↵↵ Tu↕ +24.00 (#1) · ⚡ 19.2%
  10. "7" acid administration in 73% (n↕ +23.62 (#1) · ⚡ 12.3%
  11. "7" in 1975.↵ Received↕ +23.38 (#1) · ⚡ 16.4%
  12. "7". (1973) 2↕ +23.00 (#1) · ⚡ 37.1%
  13. "7" in 1971 as a joint↕ +23.00 (#1) · ⚡ 34.6%
  14. "7" in 1976. He has↕ +22.75 (#1) · ⚡ 26.4%
  15. "7" giu 1970)↵↵An↕ +22.75 (#1) · ⚡ 41.6%
  16. "7"80, 72 S.E↕ +22.75 (#1) · ⚡ 20.0%
  17. "7"In 1876, a German↕ +22.75 (#1) · ⚡ 14.6%
  18. "7"conversion rates were 78.8%↕ +22.62 (#1) · ⚡ 14.3%
  19. "7"868 (77.0%)↕ +22.50 (#1) · ⚡ 20.0%
  20. "7" least 1976 and Hayek.↕ +22.38 (#1) · ⚡ 20.3%
Gemma 2 2B · L24/C776

Adverse health effects / medical complications

Confidence 5/5
Component profile — C776 · layer 24 · positive tail

Logit-lens tokens · positive

#1" cancer"#2" symptoms"#3" inflammation"#4" blood"#5" heart"#6" cancerous"#7" chemicals"#8" doctors"#9" stomach"#10" deadly"

R-lens tokens · positive

#1" cancer"#2" symptoms"#3" blood"#4" cancerous"#5" inflammation"#6" stomach"#7" brain"#8" cancers"#9" heart"#10" allergies"

Skewness

+2.33

Excess kurtosis

23.78rank #1357

Logcosh deviation

0.0685rank #777

Top-score occurrences · positive

  1. " from"—for example, from hearing loss after being↕ +21.38 (#1) · ⚡ 12.7%
  2. " reduce" upset stomach, and reduce fever by causing perspiration↕ +21.12 (#1) · ⚡ 11.4%
  3. " of":↵↵A history of drug or alcohol addiction↕ +20.38 (#1) · ⚡ 11.8%
  4. " with" had been recently hospitalized with pneumonia.↵↵Over↕ +19.88 (#1) · ⚡ 10.9%
  5. " of".↵↵The risk of QT prolongation may be↕ +19.88 (#1) · ⚡ 10.0%
  6. " a"↵↵After suffering from a prolonged illness, Beckett↕ +19.38 (#1) · ⚡ 8.9%
  7. " of" and has a bout of pneumonia.↵↵Share↕ +19.25 (#1) · ⚡ 11.6%
  8. " and" irregular heart beats, and congestive heart failure↕ +19.25 (#1) · ⚡ 11.7%
  9. "lowers"zes blood sugar↵↵lowers cholesterol↵↵reduces↕ +19.12 (#1) · ⚡ 9.0%
  10. " like", screens for diseases like diabetes, discusses puberty↕ +19.00 (#1) · ⚡ 9.1%
  11. " as" caused by such things as cancer cells or lowered↕ +19.00 (#1) · ⚡ 17.1%
  12. "\n\n", such as epilepsy↵↵Cardiac conditions such as↕ +18.62 (#1) · ⚡ 10.6%
  13. "," poor memory, impotence, arthritis, poor eyesight↕ +18.50 (#1) · ⚡ 13.4%
  14. " remove" operation he had to remove a fishbone in↕ +18.50 (#1) · ⚡ 8.3%
  15. " cause"Alfuzosin may cause a sudden drop in↕ +18.50 (#1) · ⚡ 8.7%
  16. " with" was in the hospital with asthma and pneumonia.↕ +18.38 (#1) · ⚡ 8.6%
  17. " and" cause allergies, asthma and irritation when the unit↕ +18.12 (#1) · ⚡ 9.4%
  18. "," for diseases like diabetes, discusses puberty and other↕ +18.00 (#1) · ⚡ 10.0%
  19. "-" way as prescription anti-diabetic medication, scientists↕ +18.00 (#2) · ⚡ 7.6%
  20. " causes"LBD], which causes extreme anxiety, delusions↕ +17.75 (#1) · ⚡ 7.8%
Gemma 2 2B · L24/C1827

Expecting "include"

Confidence 5/5
Component profile — C1827 · layer 24 · negative tail

Logit-lens tokens · negative

#1" including"#2" include"#3" includes"#4"including"#5" Including"#6"Including"#7" includ"#8"包括"#9"includes"#10" incluindo"

R-lens tokens · negative

#1" including"#2" include"#3" includes"#4"including"#5" Including"#6" includ"#7"Including"#8"包括"#9"includes"#10" Includes"

Skewness

-1.06

Excess kurtosis

7.87rank #1960

Logcosh deviation

0.0342rank #1828

Top-score occurrences · negative

  1. " here"↵↵Areas of focus here include advances in techniques↕ -17.75 (#1) · ⚡ 12.3%
  2. " groups"ent deficiency. These groups include pregnant women and↕ -16.12 (#1) · ⚡ 12.5%
  3. " services" wide range of surveying services using state of the↕ -15.81 (#1) · ⚡ 12.5%
  4. "," number of scientific areas, including exoplanets↕ -15.69 (#1) · ⚡ 16.4%
  5. " could" the house insurance contract could include:↵↵Changes↕ -15.62 (#1) · ⚡ 12.2%
  6. "," questioned by several interests, including Berkeley Patients Group↕ -15.25 (#1) · ⚡ 11.6%
  7. " which" also-rans, which also included Teenage Mutant↕ -14.81 (#1) · ⚡ 10.4%
  8. " have" residents↵Celebrity residents have included Elvis Presley,↕ -14.56 (#2) · ⚡ 7.4%
  9. " benefits". Specifically, these benefits will include the more↕ -14.56 (#1) · ⚡ 10.9%
  10. " strategies" nanomaterials. These strategies include depositing silver nanoparticles↕ -14.44 (#1) · ⚡ 10.3%
  11. ","4. Four others, including two night club↕ -14.19 (#2) · ⚡ 6.9%
  12. "," high-profile clients, including Mike Duffy.↕ -14.06 (#1) · ⚡ 11.9%
  13. "," action by various means, means that included perjury↕ -14.06 (#1) · ⚡ 8.6%
  14. "," of of land, including the island,↕ -14.00 (#1) · ⚡ 9.7%
  15. "," titles through the organization, including NAVHDA↕ -13.88 (#1) · ⚡ 10.7%
  16. " applications" and rapidly. Such applications include internal combustion engine↕ -13.81 (#1) · ⚡ 9.3%
  17. " highlights"Other first-day highlights were expected to include↕ -13.81 (#1) · ⚡ 7.3%
  18. "," variety of alternative solutions, such as a "↕ -13.81 (#1) · ⚡ 11.5%
  19. " then" "all" would then have to mean babies↕ -13.69 (#1) · ⚡ 8.5%
  20. " also" her vast experiences that also span into the likes↕ -13.62 (#1) · ⚡ 6.3%
Gemma 2 2B · L24/C2018

JavaScript DOM property/method access

Confidence 4/5
Component profile — C2018 · layer 24 · positive tail

Logit-lens tokens · positive

#1"__;"#2"sidemargin"#3"TargetException"#4" dAtA"#5"AutoresizingMask"#6" Patrimonio"#7"pathname"#8" دیکھیے"#9" {};"#10" peaking"

R-lens tokens · positive

#1"__;"#2" متعلقه"#3"sidemargin"#4"TargetException"#5"webkit"#6"ActionCreators"#7"lectricité"#8"gione"#9" coronation"#10" Somer"

Skewness

+0.07

Excess kurtosis

5.61rank #2033

Logcosh deviation

0.0239rank #2019

Top-score occurrences · positive

  1. "." {↵ document.execCommand('enable↕ +16.62 (#1) · ⚡ 5.4%
  2. "."", event.target.action);↵ ↕ +16.38 (#1) · ⚡ 6.6%
  3. "."';↵ document.body.appendChild(↕ +14.75 (#1) · ⚡ 6.2%
  4. ".";↵ document.querySelector('#containers').↕ +14.75 (#1) · ⚡ 5.2%
  5. "." {↵ document.title = data.↕ +14.44 (#1) · ⚡ 5.4%
  6. "." other things, document.GetElementsByTagName↕ +14.44 (#2) · ⚡ 5.3%
  7. "." anchorEl = document.getElementById('empty-↕ +14.31 (#1) · ⚡ 5.5%
  8. "."var swatch = document.getElementById('swat↕ +14.25 (#1) · ⚡ 5.5%
  9. "."inputName: target.name↵ }↕ +14.19 (#1) · ⚡ 4.5%
  10. ".") el = document.body;↵ ↕ +13.81 (#1) · ⚡ 4.7%
  11. "."var cur = swatch.querySelector("li.↕ +13.50 (#3) · ⚡ 4.9%
  12. "." id = body.getAttribute( "id↕ +13.31 (#1) · ⚡ 5.7%
  13. "." mainContainer = document.createElement('div');↕ +13.19 (#2) · ⚡ 5.2%
  14. ".":↵$( document.body ).on(↕ +13.19 (#1) · ⚡ 4.1%
  15. "." this:↵document.write(<div class↕ +13.00 (#1) · ⚡ 5.1%
  16. "." document.body.appendChild(mainContainer↕ +12.88 (#2) · ⚡ 5.0%
  17. "." document.body.appendChild(this.↕ +12.81 (#2) · ⚡ 4.7%
  18. "."];↵ body.innerHTML = '<div↕ +12.81 (#3) · ⚡ 4.6%
  19. "\"" value="50" />↵↵A:↕ +12.75 (#2) · ⚡ 4.0%
  20. "."selectElements = rows.getElementsByTagName("select↕ +12.75 (#2) · ⚡ 5.1%
Gemma 2 2B · L25/C1291

Anticipated future change

Confidence 4/5
Component profile — C1291 · layer 25 · positive tail

Logit-lens tokens · positive

#1" if"#2" sooner"#3" next"#4" later"#5" tomorrow"#6"."#7" before"#8" soon"#9" anytime"#10" once"

R-lens tokens · positive

#1" if"#2" sooner"#3" next"#4" later"#5" tomorrow"#6"."#7" before"#8" soon"#9" anytime"#10" once"

Skewness

+1.49

Excess kurtosis

10.05rank #1882

Logcosh deviation

0.0487rank #1292

Top-score occurrences · positive

  1. " severe" that could grow more severe in the wake of↕ +16.38 (#1) · ⚡ 13.8%
  2. " candidate" will support the GOP candidate.↵↵Contact Drew↕ +15.44 (#1) · ⚡ 12.6%
  3. " gear" kick numbers into high gear. “I think↕ +15.31 (#1) · ⚡ 13.6%
  4. " rise" those averages will likely rise in the coming weeks↕ +14.94 (#2) · ⚡ 9.3%
  5. " lower" at the pump even lower than they are today↕ +14.44 (#1) · ⚡ 11.6%
  6. " store"content could be in store over the next ↕ +14.31 (#1) · ⚡ 9.9%
  7. " schools" used in more medical schools across the United States↕ +14.31 (#1) · ⚡ 9.1%
  8. " approval" be developed for my approval, as well as↕ +14.12 (#1) · ⚡ 8.8%
  9. " sharply" but prices could drop sharply by the end of↕ +13.88 (#1) · ⚡ 10.7%
  10. " contender" could be a Calder contender next season.↵↵↕ +13.38 (#1) · ⚡ 8.3%
  11. " you" going to crumble beneath you unless you change with↕ +13.38 (#1) · ⚡ 9.2%
  12. " available"—should be commercially available in the US later↕ +13.38 (#1) · ⚡ 8.7%
  13. " addressed" that needs to be addressed forthwith.↵↵MY↕ +13.00 (#1) · ⚡ 9.1%
  14. " success" well positioned for continued success in NCAA Division II↕ +12.94 (#2) · ⚡ 6.3%
  15. " homes" likely to lose their homes,” CNN said.↕ +12.94 (#1) · ⚡ 9.6%
  16. " visible" positive effects should be visible sooner rather than later↕ +12.88 (#2) · ⚡ 5.5%
  17. " happen", a vote could happen.↕ +12.81 (#1) · ⚡ 9.4%
  18. " off" personally will be worse off four years from now↕ +12.81 (#1) · ⚡ 6.9%
  19. " rates"] to raise interest rates anytime soon," AS↕ +12.69 (#2) · ⚡ 5.2%
  20. " needs" Trump the justification he needs to fight Iran without↕ +12.62 (#2) · ⚡ 8.0%
Gemma 2 2B · L25/C1701

Salesforce development/customization

Confidence 5/5
Component profile — C1701 · layer 25 · positive tail

Logit-lens tokens · positive

#1"Apex"#2" Apex"#3"Aura"#4"apex"#5" Aura"#6" Salesforce"#7"aura"#8"Opportunity"#9"Record"#10"yntaxException"

R-lens tokens · positive

#1"Apex"#2" Apex"#3"Aura"#4"apex"#5" Aura"#6" Salesforce"#7"aura"#8"Opportunity"#9"Record"#10"yntaxException"

Skewness

+1.68

Excess kurtosis

31.30rank #1015

Logcosh deviation

0.0354rank #1702

Top-score occurrences · positive

  1. " a" prepopulate fields on a Standard layout?↵↵↕ +32.50 (#1) · ⚡ 29.9%
  2. " custom" could probably create a custom setting that holds all↕ +30.38 (#1) · ⚡ 27.5%
  3. " or" no↵ triggers or workflows will fire on↕ +30.25 (#1) · ⚡ 20.3%
  4. "s" this new record's edit view.↵↕ +30.12 (#1) · ⚡ 22.7%
  5. " custom"Outside of making a custom edit view for this↕ +29.50 (#1) · ⚡ 23.6%
  6. "public"View {↵↵ public Id propertyId;↕ +28.88 (#1) · ⚡ 22.9%
  7. " and" pre-populate standard and custom fields using this↕ +28.50 (#1) · ⚡ 19.8%
  8. " /" page, which is /xxxx.salesforce↕ +28.25 (#1) · ⚡ 19.8%
  9. " another" with a button from another org with different hard↕ +28.25 (#1) · ⚡ 25.6%
  10. " standard" pass in values for standard fields, such as↕ +28.12 (#1) · ⚡ 24.4%
  11. "public"} ↵↵ public PageReference back(){↕ +28.00 (#1) · ⚡ 24.2%
  12. "()." ApexPages.CurrentPage().getParameters().get('↕ +27.50 (#1) · ⚡ 19.4%
  13. "."sServerName + proposal.Id+editName↕ +26.62 (#1) · ⚡ 22.9%
  14. " "↵ );↵↵ insert proposal;↵↵↕ +26.25 (#1) · ⚡ 25.2%
  15. "?"0r/e?retURL=%2↕ +26.25 (#1) · ⚡ 17.1%
  16. "/"salesforce.com/a0r/↕ +25.88 (#1) · ⚡ 16.6%
  17. " uses"__c' and uses it to create a↕ +25.00 (#1) · ⚡ 22.7%
  18. "="↵ OwnerId = property.owner↕ +24.88 (#1) · ⚡ 20.4%
  19. " =";↵ pref = new PageReference(↕ +24.88 (#1) · ⚡ 21.9%
  20. "('/" = new PageReference('/' + propertyId↕ +24.75 (#1) · ⚡ 19.4%
Gemma 2 2B · L25/C1952

?

Confidence 0/5
Component profile — C1952 · layer 25 · negative tail

Logit-lens tokens · negative

#1"tonode"#2"IVEREF"#3"SourceChecksum"#4" ControllerBase"#5" itinéraires"#6"İstinadlar"#7" Warburton"#8" ProtoMessage"#9"IndentedString"#10" Arles"

R-lens tokens · negative

#1"tonode"#2"IVEREF"#3"SourceChecksum"#4" ControllerBase"#5" itinéraires"#6"İstinadlar"#7" Warburton"#8" ProtoMessage"#9"IndentedString"#10" Arles"

Skewness

-0.27

Excess kurtosis

3.52rank #2075

Logcosh deviation

0.0255rank #1953

Top-score occurrences · negative

  1. " unique", it has my unique sensibility and my reporting↕ -13.31 (#1) · ⚡ 9.1%
  2. "\n" 3.1↵|-↵! scope↕ -12.50 (#1) · ⚡ 7.3%
  3. "type" .type percent_type,@function↵↵ ↕ -12.12 (#1) · ⚡ 6.2%
  4. "," region 5b, whereby the wedge member↕ -12.06 (#1) · ⚡ 7.7%
  5. " inform" ‘Voor een vrije informatiesamenleving↕ -12.06 (#2) · ⚡ 4.6%
  6. " bills" their electricity and gas bills..........↵The profits↕ -11.69 (#3) · ⚡ 4.3%
  7. "\n" 3.1↵|- style="background↕ -11.56 (#1) · ⚡ 6.9%
  8. " Sp" the following years, Spelterini crossed the↕ -11.56 (#3) · ⚡ 5.1%
  9. " function"" />↵↵The function itself works fine,↕ -11.50 (#1) · ⚡ 4.2%
  10. "\n"2: :↵ Tk. ↕ -11.31 (#1) · ⚡ 5.8%
  11. " " source files>↵ |- CMakeLists.↕ -10.94 (#1) · ⚡ 6.2%
  12. " Sp"891, Spelterini returned to↕ -10.88 (#3) · ⚡ 5.0%
  13. "hoo" PM] cloppyhooves: ...↵↵↕ -10.75 (#3) · ⚡ 5.0%
  14. "type" .type at_type,@function↵↵ ↕ -10.62 (#2) · ⚡ 4.9%
  15. "in"6 Doğu Perinçek : % ↕ -10.50 (#2) · ⚡ 4.3%
  16. "2" (Fig. [2](#F2){↕ -10.50 (#1) · ⚡ 4.8%
  17. " B" genuine jcpenny Mario Badescu Coupon Code↕ -10.50 (#2) · ⚡ 5.1%
  18. " B" normal that each Mario Badescu Coupon Code↕ -10.44 (#2) · ⚡ 5.6%
  19. "UP"( DEL.TUPEStartDate,'↕ -10.25 (#2) · ⚡ 5.9%
  20. " B" to be type Mario Badescu Coupon Code↕ -10.19 (#2) · ⚡ 5.1%
Qwen 3.5 9B Base · L0/C609

Store

Confidence 5/5
Component profile — C609 · layer 0 · negative tail

Logit-lens tokens · negative

#1"/store"#2".store"#3" store"#4" Store"#5"(store"#6"store"#7" STORE"#8"upt"#9".Store"#10"ju"

R-lens tokens · negative

#1"(ST"#2"(store"#3" createStore"#4"(SS"#5"/store"#6" STORE"#7".Store"#8"hips"#9"ystore"#10"peicher"

Skewness

-29.79

Excess kurtosis

1593.71rank #481

Logcosh deviation

0.2062rank #610

Top-score occurrences · negative

  1. " Store"↵import type { Store } from 'redux↕ -71.50 (#1) · ⚡ 45.1%
  2. "Store"<title>IDBObjectStore.createIndex() -↕ -69.50 (#1) · ⚡ 46.1%
  3. " store"↵});↵↵const store = createStore(rootReducer↕ -68.50 (#1) · ⚡ 61.7%
  4. " store" Persistor↵ store: Store↵}↕ -67.00 (#1) · ⚡ 62.1%
  5. "store" doing this.get('store') inside any route↕ -66.50 (#1) · ⚡ 50.0%
  6. "Store" "Deletes a androidStoreApp."↵author↕ -66.00 (#1) · ⚡ 40.2%
  7. " store" make an API call store email details in DB↕ -66.00 (#1) · ⚡ 69.9%
  8. "store"keyalg RSA -storetype PKCS1↕ -65.00 (#1) · ⚡ 43.4%
  9. " Store" {↵ Store (0, Index↕ -65.00 (#1) · ⚡ 53.1%
  10. "store"une-apps-androidstoreapp.md).↵↵↕ -64.50 (#1) · ⚡ 41.6%
  11. " Store".↵↵Powerfull Store Management↵↵Extensions &↕ -64.00 (#1) · ⚡ 52.3%
  12. " store" of the Archos store it says the ↕ -63.00 (#1) · ⚡ 70.3%
  13. "Store".Register<MockDataStore>();↵ MainPage↕ -61.50 (#1) · ⚡ 41.8%
  14. " store" this in the app store?↵--------------------------------↵↵↕ -61.25 (#1) · ⚡ 71.5%
  15. " store".↵↵The three storey-high building was↕ -61.00 (#1) · ⚡ 64.8%
  16. " store" public actual inline fun storeAt(index: Int↕ -61.00 (#1) · ⚡ 62.1%
  17. "Store"::BareCryptoStorePtr;↵use↕ -61.00 (#1) · ⚡ 41.0%
  18. " store" access to a data store. In particular,↕ -61.00 (#1) · ⚡ 62.1%
  19. " store" as an argument or store it one of two↕ -60.75 (#1) · ⚡ 71.5%
  20. "-store"↵server.ssl.key-store=/etc/lets↕ -60.75 (#1) · ⚡ 28.9%
Qwen 3.5 9B Base · L0/C3472

Promise

Confidence 5/5
Component profile — C3472 · layer 0 · negative tail

Logit-lens tokens · negative

#1"盖世"#2"wechsel"#3"ListOf"#4"ﾉ"#5"甚麼"#6" Tiên"#7"磯"#8"古屋"#9" Vare"#10"不可及"

R-lens tokens · negative

#1" endeavour"#2" moreover"#3" undertaking"#4" akin"#5" envis"#6" undertaken"#7" nonetheless"#8" meanwhile"#9" scarcely"#10" harbour"

Skewness

-1.53

Excess kurtosis

60.16rank #3085

Logcosh deviation

0.0807rank #3473

Top-score occurrences · negative

  1. "Promise"use GuzzleHttp\Promise\PromiseInterface;↕ -37.50 (#2) · ⚡ 5.7%
  2. " Promise" = x => new Promise(res => setTimeout(res↕ -37.00 (#2) · ⚡ 5.2%
  3. " promise"yResults = await promiseTimeout(↵ ↕ -36.00 (#2) · ⚡ 9.5%
  4. " Promise".↵ at Promise (C:\Image↕ -35.75 (#2) · ⚡ 7.4%
  5. " promise"(request) using the promise when returned to stop↕ -35.25 (#2) · ⚡ 10.5%
  6. " promise" the series was the promise of ‘annual updates↕ -34.00 (#2) · ⚡ 11.3%
  7. " Promise"’s Xsan and Promise RAID and Allows X↕ -33.00 (#2) · ⚡ 8.1%
  8. " promise", the discounts only promise to get better (↕ -32.50 (#2) · ⚡ 11.4%
  9. " promise" {Promise} A promise of the view.↕ -31.88 (#2) · ⚡ 9.7%
  10. " promise" asked about Dix's promise to freeze ferry fares↕ -31.75 (#2) · ⚡ 10.8%
  11. " promise" will not keep its promise to bring ridesh↕ -31.25 (#2) · ⚡ 11.4%
  12. " promise" are late. This promise can save you up↕ -31.12 (#2) · ⚡ 10.4%
  13. " promise"So I made a promise to myself↵To↕ -31.00 (#2) · ⚡ 10.5%
  14. " promise" failed to keep its promise to replace council tax↕ -30.38 (#2) · ⚡ 10.9%
  15. " promise". Trump used the promise of a role in↕ -30.25 (#2) · ⚡ 10.7%
  16. " promise" died I had to promise him to give you↕ -29.88 (#2) · ⚡ 10.3%
  17. " promise" this study demonstrates the promise of Lorentz↕ -29.88 (#2) · ⚡ 8.9%
  18. " Promise" Array, not a Promise,↵↵All async↕ -29.75 (#2) · ⚡ 7.8%
  19. " Promise" {↵ return Promise.all(self.products.map↕ -29.62 (#2) · ⚡ 6.2%
  20. " promise" and promotions. We promise you wont be bomb↕ -29.62 (#2) · ⚡ 9.2%
Qwen 3.5 9B Base · L1/C3036

Do something / nothing

Confidence 5/5
Component profile — C3036 · layer 1 · negative tail

Logit-lens tokens · negative

#1"卿"#2"不到的"#3" bigger"#4"ittle"#5"正视"#6"actical"#7"趁"#8"化学反应"#9"tn"#10"RAL"

R-lens tokens · negative

#1"s"#2"ness"#3"re"#4"sob"#5"ss"#6"soever"#7"sWith"#8"shire"#9"nya"#10"ted"

Skewness

-0.79

Excess kurtosis

41.92rank #3206

Logcosh deviation

0.0701rank #3037

Top-score occurrences · negative

  1. " nothing" Galatzan did nothing to block the closure↕ -31.38 (#2) · ⚡ 9.3%
  2. " nothing" away choices and does nothing to solve the underlying↕ -31.25 (#2) · ⚡ 9.7%
  3. " something"↵↵ //Do something with the object↵↕ -30.12 (#3) · ⚡ 8.9%
  4. " nothing"A day off did nothing to slow down Curtis↕ -29.75 (#2) · ⚡ 7.6%
  5. " anything"'t have to do anything↵#define BOOST_SER↕ -27.00 (#2) · ⚡ 10.6%
  6. " nothing" The second server does nothing on the database,↕ -27.00 (#2) · ⚡ 9.1%
  7. "THING"ING TO SAY ANYTHING TO GET ELECTED↕ -27.00 (#1) · ⚡ 5.2%
  8. " anything" and doesn't do anything, do you get↕ -26.75 (#2) · ⚡ 11.0%
  9. " anything" thought we could do anything about violent crime.↕ -26.75 (#2) · ⚡ 11.1%
  10. " anything"creation, to do anything substantial to help the↕ -26.38 (#2) · ⚡ 11.4%
  11. " nothing" default canceller does nothing↵ conn.cancel↕ -26.25 (#2) · ⚡ 7.9%
  12. "Something" public void doSomething() {↵ ↕ -26.25 (#2) · ⚡ 6.3%
  13. " something" United States to do something about the Israeli-P↕ -26.12 (#3) · ⚡ 12.7%
  14. " anything" I wouldn't do anything. There may↕ -26.12 (#3) · ⚡ 10.1%
  15. " anything" it hasn't done anything since. Is this↕ -26.00 (#3) · ⚡ 10.3%
  16. " something" see a pastor do something about the problems with↕ -25.75 (#3) · ⚡ 10.8%
  17. " nothing" them as such does nothing to prove it to↕ -25.75 (#2) · ⚡ 7.9%
  18. " anything" relegation would do anything at all. The↕ -25.50 (#2) · ⚡ 11.3%
  19. " anything" no need to do anything to the rotor when↕ -25.50 (#2) · ⚡ 10.7%
  20. " something" be able to do something in my plugin like↕ -25.38 (#3) · ⚡ 11.5%
Qwen 3.5 9B Base · L1/C3298

spokesman

Confidence 5/5
Component profile — C3298 · layer 1 · positive tail

Logit-lens tokens · positive

#1"er"#2"白金"#3"�"#4"移"#5"erged"#6"联"#7"vara"#8"endorf"#9"以身"#10"ería"

R-lens tokens · positive

#1" booze"#2"்"#3" staffers"#4"iée"#5"iés"#6"ي"#7"esas"#8"-esque"#9" spokesman"#10" onstage"

Skewness

+0.50

Excess kurtosis

53.94rank #3009

Logcosh deviation

0.0629rank #3299

Top-score occurrences · positive

  1. " spokesman" money.↵↵A spokesman for Mayor Michael Bloomberg↕ +37.75 (#1) · ⚡ 16.2%
  2. " spokesman" missing.↵↵A spokesman for the Metropolitan Police↕ +37.50 (#1) · ⚡ 16.0%
  3. " spokesman" communities.'↵↵A spokesman from Traveller Movement↕ +36.00 (#1) · ⚡ 16.7%
  4. " spokesman"ettenham, a spokesman for the museum of↕ +35.50 (#1) · ⚡ 17.7%
  5. " spokesman", Chinese Foreign Ministry spokesman Geng Shuang↕ +34.75 (#1) · ⚡ 16.4%
  6. " spokesman" the Commons," a spokesman said.↵↵But↕ +34.50 (#1) · ⚡ 18.3%
  7. " spokesman", said state fire spokesman Dennis Mathisen.↕ +34.50 (#1) · ⚡ 15.8%
  8. " spokesman".↵↵A government spokesman said "The published↕ +34.00 (#1) · ⚡ 16.5%
  9. " spokesman"stry and Fire Protection spokesman Daniel Berlant,↕ +34.00 (#1) · ⚡ 15.6%
  10. " spokesman" said an Ofcom spokesman.↵↵'Having↕ +34.00 (#1) · ⚡ 17.1%
  11. " spokesman"," Federal Communications Commission spokesman Neil Grace said.↕ +33.75 (#1) · ⚡ 15.8%
  12. " spokesman".↵↵An FA spokesman said: 'The↕ +33.75 (#1) · ⚡ 17.2%
  13. " spokesman"esneau, a spokesman for the Prime Minister↕ +33.50 (#1) · ⚡ 17.4%
  14. " spokesman" reservoir, said fire spokesman Matthew Chambers.↵↵↕ +33.50 (#1) · ⚡ 16.2%
  15. " spokesman" James Garrow, spokesman for the Philadelphia Department↕ +32.75 (#1) · ⚡ 15.3%
  16. " spokesman"6]↵↵The spokesman for the Shanti↕ +32.00 (#1) · ⚡ 13.8%
  17. " spokesman" John Lowe A police spokesman said: "The↕ +32.00 (#1) · ⚡ 16.1%
  18. " spokesman".↵↵Regional police spokesman Romeo Galgo said↕ +32.00 (#1) · ⚡ 16.6%
  19. " spokesman", a Thune spokesman, told National Journal↕ +31.50 (#1) · ⚡ 15.8%
  20. " spokesman" false claim, campaign spokesman Andrew Bates insisted that↕ +31.38 (#1) · ⚡ 15.7%
Qwen 3.5 9B Base · L2/C231

User

Confidence 5/5
Component profile — C231 · layer 2 · negative tail

Logit-lens tokens · negative

#1"oggio"#2"/user"#3"/student"#4"izzo"#5"/User"#6"eroy"#7"adio"#8"/customer"#9"(user"#10"-user"

R-lens tokens · negative

#1"[user"#2"(user"#3"(User"#4"=user"#5"/user"#6"(USER"#7":user"#8" setUser"#9" updateUser"#10" USER"

Skewness

-18.70

Excess kurtosis

666.42rank #783

Logcosh deviation

0.1858rank #232

Top-score occurrences · negative

  1. "_USER" alternative to GWL_USERDATA for storing an↕ -54.50 (#1) · ⚡ 25.0%
  2. " user"I have a table user_comission_configuration_history↕ -53.50 (#1) · ⚡ 46.1%
  3. "USER"<|endoftext|>Related Articles↵↵USER COMMENTS↵↵Posted by↕ -50.50 (#1) · ⚡ 33.0%
  4. " user" including product design, user-centred design,↕ -50.50 (#1) · ⚡ 53.1%
  5. " user"11 @@↵ userdom_search_user_home↕ -47.00 (#1) · ⚡ 41.4%
  6. " user" cannot be used in user SQL directly.↵↵↕ -47.00 (#1) · ⚡ 61.7%
  7. "User" unsigned int m_uiUserFlag;↵ ↕ -47.00 (#1) · ⚡ 34.2%
  8. "User" Campaign Management Platforms↵User behavior data video ↕ -47.00 (#1) · ⚡ 49.2%
  9. " user" in, but creating user defined functions, including↕ -46.50 (#1) · ⚡ 61.3%
  10. " User" "select User_name, User_Se↕ -46.00 (#1) · ⚡ 33.0%
  11. " USER" this:↵| USER | MANAGER↕ -46.00 (#1) · ⚡ 31.8%
  12. " User" Using the HRS User Guides and the RAND↕ -45.25 (#1) · ⚡ 41.6%
  13. " user"se - write user defined value to the↕ -45.25 (#1) · ⚡ 53.5%
  14. "User"Taxonomy.↵User.↵Field API↕ -44.75 (#1) · ⚡ 61.7%
  15. "User" // "User Id={2};↕ -44.50 (#1) · ⚡ 44.9%
  16. " User" NOTE TO USER: User expressly acknowledges and agrees↕ -44.25 (#1) · ⚡ 38.5%
  17. " user"2) study the user interface requirements of a↕ -44.25 (#1) · ⚡ 53.5%
  18. " user" to be queried for user or process data.↕ -44.25 (#1) · ⚡ 66.4%
  19. "R"psr = USR_MODE; ↕ -44.00 (#1) · ⚡ 8.7%
  20. " Users" have a Model of Users and a Model of↕ -44.00 (#1) · ⚡ 37.3%
Qwen 3.5 9B Base · L2/C2327

---

Confidence 5/5
Component profile — C2327 · layer 2 · positive tail

Logit-lens tokens · positive

#1"هِ"#2".decorate"#3" Coro"#4"ri"#5"re"#6"POS"#7"->"#8"_ENTRIES"#9"这时候"#10"akk"

R-lens tokens · positive

#1"--"#2"!--"#3"\"--"#4" \"--"#5")--"#6".--"#7"/--"#8" --"#9"---"#10"(\"--"

Skewness

+4.17

Excess kurtosis

96.64rank #2519

Logcosh deviation

0.0899rank #2328

Top-score occurrences · positive

  1. "---"<|endoftext|>---↵"Missing document↕ +34.50 (#1) · ⚡ 13.2%
  2. "---"<|endoftext|>--- contrib/virt.te↕ +34.50 (#1) · ⚡ 13.2%
  3. "---"<|endoftext|>---↵exp_name:↕ +34.50 (#1) · ⚡ 13.2%
  4. "---"<|endoftext|>---↵layout: example↕ +34.50 (#1) · ⚡ 13.2%
  5. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  6. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  7. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  8. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  9. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  10. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  11. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  12. "---"<|endoftext|>---↵author:↵↕ +34.25 (#1) · ⚡ 13.1%
  13. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  14. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  15. "---"<|endoftext|>---↵abstract: '↕ +34.25 (#1) · ⚡ 13.1%
  16. "---"<|endoftext|>---↵abstract: |↕ +34.25 (#1) · ⚡ 13.1%
  17. "---"<|endoftext|>---↵abstract: |↕ +34.25 (#1) · ⚡ 13.1%
  18. "--"<|endoftext|>#-- copyright↵# Open↕ +32.75 (#1) · ⚡ 6.3%
  19. "--"<|endoftext|>PARIS--French anti-terror↕ +32.25 (#1) · ⚡ 10.2%
  20. "---"notes # Lotus↵---↵Run Lotus Notes↕ +31.50 (#1) · ⚡ 15.0%
Qwen 3.5 9B Base · L2/C3057

-y, -ie nickname

Confidence 1/5
Component profile — C3057 · layer 2 · positive tail

Logit-lens tokens · positive

#1"тр"#2"多孔"#3"Flo"#4" Mass"#5"ddl"#6".ng"#7"瀑"#8"不高兴"#9" ng"#10"oura"

R-lens tokens · positive

#1" gonna"#2" meistens"#3"yyy"#4" cuz"#5"itary"#6" zufolge"#7"hood"#8" Nowadays"#9"colar"#10"ields"

Skewness

+0.30

Excess kurtosis

25.22rank #3509

Logcosh deviation

0.0689rank #3058

Top-score occurrences · positive

  1. "ie"0 PM] sammiepie: which family↕ +33.00 (#1) · ⚡ 3.2%
  2. "ie"1 PM] sammiepie: ^↵↵↕ +26.75 (#1) · ⚡ 3.6%
  3. "nie"~~~↵JonnieCache↵Yeah,↕ +21.12 (#1) · ⚡ 1.9%
  4. "Johnny"<|endoftext|>Johnny Rodriguez↵↵Juan R↕ +21.00 (#1) · ⚡ 3.1%
  5. "nie"ellant.↵Johnnie L. Caldwell,↕ +20.62 (#1) · ⚡ 2.9%
  6. "ie"1) for Henie Onstad Kunsts↕ +20.62 (#1) · ⚡ 1.2%
  7. "vie" the wall of Stevie Wonder circa 1↕ +20.25 (#3) · ⚡ 1.6%
  8. "ie"11, Henie Onstad Kunsts↕ +19.75 (#1) · ⚡ 1.3%
  9. "asp" etc and even .asp.net applications for web↕ +19.62 (#4) · ⚡ 2.0%
  10. "ddie"... More↵↵Eddie Cusack (↕ +18.38 (#2) · ⚡ 2.5%
  11. " Chamber"., and collaborated with Chamber Orchestra Kremlin, Ensemble↕ +18.12 (#3) · ⚡ 2.8%
  12. "ie" began. “Frankie Kazarian.”↵↵↕ +18.00 (#3) · ⚡ 2.6%
  13. " Chamber"obineau, Chamberlain) and its↕ +17.62 (#2) · ⚡ 3.1%
  14. "ie" that have made Winnie Mandela such a controversial↕ +17.50 (#1) · ⚡ 1.2%
  15. "ny" Lions left back Donny Toia said.↕ +17.38 (#1) · ⚡ 1.6%
  16. "y"]\]. Chalky areas in rice grains↕ +17.12 (#2) · ⚡ 1.3%
  17. "asp" href="http://asp.net" rel="↕ +17.12 (#4) · ⚡ 1.6%
  18. " ASP" from generic repository in ASP.NET controller class?↕ +17.00 (#3) · ⚡ 2.8%
  19. "ie"External links↵Willie Mataka at West↕ +16.88 (#1) · ⚡ 2.4%
  20. " Chamber"), to Houston Stewart Chamberlain’s disquisition↕ +16.88 (#2) · ⚡ 2.5%
Qwen 3.5 9B Base · L4/C565

Week / month

Confidence 4/5
Component profile — C565 · layer 4 · negative tail

Logit-lens tokens · negative

#1"/month"#2"-month"#3"-week"#4"-long"#5"week"#6" month"#7"ender"#8"ets"#9"ires"#10"ืด"

R-lens tokens · negative

#1" fortn"#2"/month"#3" month"#4" week"#5" Month"#6" WEEK"#7"-week"#8" Week"#9" weeks"#10"week"

Skewness

-9.49

Excess kurtosis

276.59rank #676

Logcosh deviation

0.1232rank #566

Top-score occurrences · negative

  1. "apt". Toata saptamana viitoare lip↕ -53.00 (#1) · ⚡ 12.2%
  2. " Week"<|endoftext|>↵↵This Week in HN podcast –↕ -52.00 (#1) · ⚡ 19.6%
  3. "EEK"WHERE DATEPART(WEEK, OPCH.Doc↕ -48.50 (#1) · ⚡ 18.9%
  4. " week"<|endoftext|>A week ago we published a↕ -45.50 (#1) · ⚡ 30.1%
  5. "EEK"AUTICAL^WEEKLY IN THE^↕ -45.25 (#1) · ⚡ 24.0%
  6. "EEK" Ill. (WEEK) — This morning↕ -45.25 (#1) · ⚡ 20.6%
  7. " week"13↵↵This week's edition of "↕ -43.00 (#1) · ⚡ 25.2%
  8. "amana" Toata saptamana viitoare lipse↕ -42.00 (#1) · ⚡ 19.2%
  9. " week" equip...↵↵This week, allegations started flying↕ -41.50 (#1) · ⚡ 27.1%
  10. "j"ine↵Nedeljnik↵Newsweek↕ -40.75 (#1) · ⚡ 7.4%
  11. "ikon"ämästä: viikonlopuista,↕ -40.50 (#2) · ⚡ 12.1%
  12. " Week"And Now, a Week of Full-On Terrorism↕ -40.25 (#1) · ⚡ 23.0%
  13. "EEK"↵ "WEEKENDRANGE":↕ -39.75 (#1) · ⚡ 16.5%
  14. "ck"ligare flera veckor.↵↵Tre↕ -39.50 (#1) · ⚡ 20.7%
  15. " week" euthanised at week 1, ↕ -39.25 (#1) · ⚡ 21.0%
  16. " week" poly family! This week it’s my turn↕ -38.75 (#1) · ⚡ 23.0%
  17. " week"les.↵↵A week before Election Day in↕ -38.25 (#1) · ⚡ 28.5%
  18. "week"↵ , extract(week from registration_time)%↕ -38.00 (#1) · ⚡ 18.6%
  19. " week"'s it for this week! If you like↕ -37.25 (#1) · ⚡ 27.0%
  20. " week" de Abril – a week-long fiesta of par↕ -37.25 (#1) · ⚡ 29.7%
Qwen 3.5 9B Base · L4/C1156

Introduction

Confidence 5/5
Component profile — C1156 · layer 4 · negative tail

Logit-lens tokens · negative

#1"ductory"#2"repid"#3" introduction"#4"egn"#5"介绍"#6"RODUCTION"#7"Introduction"#8" introducir"#9" introd"#10" Introdu"

R-lens tokens · negative

#1" emphas"#2" initialise"#3" Whilst"#4"ductory"#5" personalised"#6"Whilst"#7" specialised"#8" recognised"#9" programme"#10" Initialise"

Skewness

-11.61

Excess kurtosis

618.13rank #117

Logcosh deviation

0.0998rank #1157

Top-score occurrences · negative

  1. "Introduction"'↵---↵↵Introduction. {#sec↕ -67.00 (#1) · ⚡ 34.0%
  2. "Introduction"'↵---↵↵Introduction {#sec:int↕ -66.50 (#1) · ⚡ 30.7%
  3. "Introduction".'↵---↵↵Introduction↵============↵↵T↕ -66.00 (#1) · ⚡ 32.2%
  4. "Introduction"frared↵---↵↵Introduction↵============↵↵Many↕ -65.50 (#1) · ⚡ 34.2%
  5. "Introduction"'↵---↵↵Introduction↵============↵↵Three↕ -65.00 (#1) · ⚡ 32.0%
  6. "Introduction" low complexity.↵↵Introduction↵============↵↵Bi↕ -65.00 (#1) · ⚡ 38.1%
  7. "Introduction"]'↵---↵↵Introduction {#sec:int↕ -63.75 (#1) · ⚡ 36.3%
  8. "Introduction" diseases↵---↵↵Introduction↵============↵↵The↕ -63.75 (#1) · ⚡ 35.5%
  9. "Introduction"forward rapidity↵↵Introduction {#Intro}↕ -63.50 (#1) · ⚡ 36.1%
  10. "Introduction"aser\]↵↵Introduction↵============↵↵Recently↕ -63.00 (#1) · ⚡ 40.0%
  11. "Introduction" parallel computation.↵↵Introduction\[sec:introduction↕ -62.25 (#1) · ⚡ 34.4%
  12. "Introduction"'↵---↵↵Introduction↵============↵↵In↕ -61.75 (#1) · ⚡ 28.1%
  13. "Introduction" ,correlation↵↵Introduction {#sec:int↕ -61.50 (#1) · ⚡ 36.9%
  14. "Introduction" band.]{}↵↵Introduction↵============↵↵A↕ -61.25 (#1) · ⚡ 32.0%
  15. "RODUCTION"<|endoftext|>INTRODUCTION {#s1↕ -60.75 (#1) · ⚡ 45.7%
  16. "RODUCTION"<|endoftext|>INTRODUCTION {#sec1↕ -60.75 (#1) · ⚡ 45.7%
  17. "RODUCTION"<|endoftext|>INTRODUCTION {#sec1↕ -60.75 (#1) · ⚡ 45.7%
  18. "Introduction"<|endoftext|>Introduction↵============↵↵Ag↕ -60.00 (#1) · ⚡ 53.1%
  19. "Introduction"<|endoftext|>Introduction↵============↵↵Now↕ -60.00 (#1) · ⚡ 53.1%
  20. "Introduction"<|endoftext|>Introduction↵============↵↵The↕ -60.00 (#1) · ⚡ 53.1%
Qwen 3.5 9B Base · L4/C2219

Measurement/unit-conversion math question

Confidence 4/5
Component profile — C2219 · layer 4 · negative tail

Logit-lens tokens · negative

#1"귀"#2"返回首页"#3"回帖"#4"查看答案"#5"看一眼"#6"edList"#7"下载使用"#8" goog"#9"圆圈"#10"xFB"

R-lens tokens · negative

#1" recieved"#2" seperate"#3" apologise"#4" Whilst"#5"utilise"#6" seper"#7"Whilst"#8" Maths"#9" preventative"#10" maths"

Skewness

-2.51

Excess kurtosis

34.17rank #2919

Logcosh deviation

0.0728rank #2220

Top-score occurrences · negative

  1. " many"29↵How many centimeters are there↕ -25.50 (#1) · ⚡ 9.4%
  2. " many"50↵How many micrometers are↕ -25.00 (#1) · ⚡ 9.2%
  3. " many"26↵How many seconds are there in↕ -24.38 (#1) · ⚡ 9.1%
  4. " "0↵What is 0.53↕ -22.38 (#1) · ⚡ 11.1%
  5. " rounded"6794 rounded to the nearest ↕ -21.62 (#2) · ⚡ 4.4%
  6. " of"3/25 of a milligram in↕ -21.62 (#1) · ⚡ 7.6%
  7. " kilograms"9↵How many kilograms are there in ↕ -21.62 (#1) · ⚡ 6.2%
  8. " minutes"0793 minutes in seconds?↵↕ -20.88 (#2) · ⚡ 7.2%
  9. " days"2215 days?↵84↕ -20.75 (#2) · ⚡ 6.3%
  10. "mg"5582mg to grams.↵↕ -20.75 (#2) · ⚡ 5.1%
  11. " "8↵What is 53/5↕ -20.75 (#1) · ⚡ 14.1%
  12. " "5↵What is 1790↕ -20.62 (#1) · ⚡ 14.0%
  13. " many"80↵How many micrometers are↕ -20.50 (#3) · ⚡ 5.5%
  14. " " seconds are there in 2629↕ -20.38 (#1) · ⚡ 14.5%
  15. " "0↵What is 8835↕ -20.38 (#1) · ⚡ 14.3%
  16. "imeters"531 centimeters?↵75↕ -20.25 (#2) · ⚡ 5.6%
  17. " rounded"63.1 rounded to the nearest ↕ -20.25 (#3) · ⚡ 4.0%
  18. " of"53/5 of a gram in mill↕ -20.12 (#1) · ⚡ 8.2%
  19. " " kilograms are there in 342.↕ -20.00 (#1) · ⚡ 17.0%
  20. " in" many grams are there in twenty-one fifths↕ -19.88 (#1) · ⚡ 8.3%
Qwen 3.5 9B Base · L5/C3133

Diabetes

Confidence 5/5
Component profile — C3133 · layer 5 · negative tail

Logit-lens tokens · negative

#1"yg"#2"�"#3"XM"#4"aglio"#5"前述"#6" aforementioned"#7"igin"#8" Sugar"#9"duc"#10"d"

R-lens tokens · negative

#1"ereum"#2" Diabetes"#3" SHALL"#4"alytics"#5"糖"#6"يض"#7" диабет"#8"蔗糖"#9"杂谈"#10"cript"

Skewness

-0.67

Excess kurtosis

16.44rank #3344

Logcosh deviation

0.0471rank #3134

Top-score occurrences · negative

  1. " diabetic" associated with severity of diabetic retinopathy.↕ -27.12 (#2) · ⚡ 12.5%
  2. "abetes"1].↵↵Diabetes is often preceded and↕ -25.12 (#2) · ⚡ 11.9%
  3. " diabetes") among people with diabetes. Synovial↕ -24.88 (#1) · ⚡ 12.9%
  4. " diabetes" and information seeking for diabetes care among Black Caribbean↕ -24.75 (#2) · ⚡ 10.8%
  5. " diabetes"%. High BMI, diabetes, high blood pressure↕ -23.88 (#1) · ⚡ 12.7%
  6. " diabetic", podcaster, diabetic, toy collector,↕ -23.38 (#1) · ⚡ 8.1%
  7. "DM"abetes (pre-DM), typically defined as↕ -23.12 (#1) · ⚡ 5.2%
  8. " diabetes" a sole treatment for diabetes and used with other↕ -22.62 (#2) · ⚡ 13.6%
  9. " Diabetes"↵↵Lifestyle, Diabetes, and Cardiovascular↕ -22.38 (#2) · ⚡ 10.6%
  10. " Diabetes" With Type 2 Diabetes: A Mixed Methods↕ -21.88 (#2) · ⚡ 9.2%
  11. " diabetes" against heart disease, diabetes, and perhaps even↕ -21.88 (#1) · ⚡ 10.7%
  12. " diabetes", nephritis, diabetes, arthritis, fib↕ -21.75 (#1) · ⚡ 13.9%
  13. "DM" (HTN/DM) amongst patients with↕ -21.38 (#2) · ⚡ 7.4%
  14. "abetes" including obesity/diabetes, cardiovascular disorders,↕ -21.25 (#2) · ⚡ 12.2%
  15. " diabetes" patients), Type II diabetes mellitus (2↕ -21.00 (#2) · ⚡ 10.7%
  16. " diabetes" for developing obesity and diabetes.↵↵Anyone can↕ -20.75 (#1) · ⚡ 13.4%
  17. " diabetic" center-involving diabetic macular edema↕ -20.75 (#1) · ⚡ 11.1%
  18. " diabetic" immunity in the spontaneously diabetic BB rat.↵↕ -20.62 (#2) · ⚡ 7.1%
  19. " diabetes" findings were indicative of diabetes insipidus↕ -20.62 (#1) · ⚡ 11.3%
  20. " diabetic" associated with severity of diabetic retinopathy.↕ -20.50 (#2) · ⚡ 11.1%
Qwen 3.5 9B Base · L6/C269

Figurative expression / metaphor

Confidence 5/5
Component profile — C269 · layer 6 · negative tail

Logit-lens tokens · negative

#1"般的"#2"般"#3"般地"#4"似的"#5"式的"#6"[image"#7"万丈"#8" gord"#9"一样的"#10" metaphor"

R-lens tokens · negative

#1"-esque"#2" folks"#3" stuff"#4"的事儿"#5" goodies"#6" mighty"#7" metaphor"#8"-ish"#9" booze"#10"这事儿"

Skewness

-6.82

Excess kurtosis

127.03rank #1422

Logcosh deviation

0.1292rank #270

Top-score occurrences · negative

  1. " extingu" analytics is a fire extinguisher. Your website↕ -49.50 (#1) · ⚡ 15.9%
  2. " driven" is a car being driven by Trump and Rudy↕ -43.50 (#1) · ⚡ 16.5%
  3. " home" climate-change chickens coming home to roost.↕ -38.50 (#1) · ⚡ 12.4%
  4. " fork" is like a tuning fork realigning us↕ -38.25 (#1) · ⚡ 10.2%
  5. "udder" here, the rudder is clearly well-held↕ -37.25 (#1) · ⚡ 11.4%
  6. " ump" as if an economic umpire had blown the↕ -37.00 (#1) · ⚡ 13.4%
  7. " athlete" a well-honed athlete. A new optional↕ -37.00 (#1) · ⚡ 16.1%
  8. "bu" it were a lifebuoy when you should↕ -36.75 (#1) · ⚡ 8.9%
  9. " nose" just cutting off your nose to spite your face↕ -36.00 (#3) · ⚡ 6.8%
  10. " in" needs to kick sand in the face of the↕ -35.50 (#1) · ⚡ 9.0%
  11. " playing" competitors from the economic playing field to recuperate↕ -35.25 (#1) · ⚡ 19.7%
  12. " balloon" inflate, but the balloon will always burst.’↕ -33.50 (#1) · ⚡ 10.0%
  13. " shot" has fired a warning shot with its decision to↕ -33.25 (#1) · ⚡ 8.7%
  14. " gun" region. The starting gun for Latin American startups↕ -33.00 (#1) · ⚡ 9.2%
  15. " grass" into the 'long grass' by diversionary↕ -32.50 (#1) · ⚡ 7.3%
  16. " rabbits" also lead down the rabbits hole of looking at↕ -32.00 (#1) · ⚡ 6.8%
  17. " a" “pushing on a string”.↕ -31.75 (#2) · ⚡ 9.1%
  18. " hole"camel through the needle hole” than get noticed↕ -31.50 (#1) · ⚡ 9.2%
  19. " canción" voces en una común canción" (Caputo↕ -31.50 (#1) · ⚡ 12.5%
  20. " driver" the hit-and-run driver berating the j↕ -31.12 (#1) · ⚡ 14.1%
Qwen 3.5 9B Base · L6/C609

Data

Confidence 5/5
Component profile — C609 · layer 6 · negative tail

Logit-lens tokens · negative

#1" liệu"#2"/data"#3"=data"#4"-testid"#5"这些数据"#6"(Data"#7"-data"#8"illons"#9"uality"#10"/Data"

R-lens tokens · negative

#1"这些数据"#2"=data"#3"/data"#4" data"#5"(Data"#6"(data"#7",data"#8"-data"#9" Data"#10"[data"

Skewness

-6.85

Excess kurtosis

146.05rank #1202

Logcosh deviation

0.1086rank #610

Top-score occurrences · negative

  1. " data". Bizopps data leads may be the↕ -34.00 (#1) · ⚡ 13.8%
  2. " data" gaming to GPUs for data centers. Those products↕ -33.25 (#1) · ⚡ 18.6%
  3. " data"/bin↵↵# Move data to staging folder↵↕ -33.00 (#1) · ⚡ 19.1%
  4. " data"ing, why does data appear in the directory↕ -32.75 (#1) · ⚡ 23.4%
  5. " data"<|endoftext|>// This data is machine generated from↕ -31.88 (#1) · ⚡ 20.3%
  6. "Data"<|endoftext|>Data in MongoDB has a↕ -31.25 (#1) · ⚡ 27.1%
  7. " data" the ways in which data plays an important role↕ -30.62 (#1) · ⚡ 21.9%
  8. " data" found increasing adoption in data center applications since a↕ -30.50 (#1) · ⚡ 18.4%
  9. " data"ungry applications and data transmitting between sites,↕ -30.50 (#1) · ⚡ 24.9%
  10. " data" attempting to compare my data, which is in↕ -30.25 (#1) · ⚡ 26.8%
  11. "Data" maintains state.↵Data symbols are the typical↕ -30.25 (#1) · ⚡ 25.4%
  12. "Data" for simplicity.↵↵Data driven testing↵↵An↕ -30.00 (#1) · ⚡ 26.4%
  13. "-data"fload/fontloader-data-con.lua↵share↕ -29.88 (#1) · ⚡ 17.9%
  14. " data" to some of the data science projects being developed↕ -29.50 (#1) · ⚡ 21.7%
  15. " data", cloud computing, data analysis. In security↕ -29.50 (#1) · ⚡ 22.2%
  16. "Data"-Terabyte-Data-Capacity Optical↕ -29.50 (#1) · ⚡ 15.7%
  17. " data" CDMA system, data to be transmitted are↕ -29.25 (#1) · ⚡ 16.6%
  18. "_DATA"4↵ AF_DATAKIT = ↕ -29.12 (#1) · ⚡ 14.3%
  19. " data" present invention relates to data processing by digital computer↕ -29.00 (#1) · ⚡ 23.5%
  20. " data" allows the transfer of data (such as audio↕ -29.00 (#1) · ⚡ 23.1%
Qwen 3.5 9B Base · L6/C992

Notice / note as information

Confidence 5/5
Component profile — C992 · layer 6 · positive tail

Logit-lens tokens · positive

#1" note"#2" Note"#3"note"#4"注意"#5"注意的是"#6"Note"#7">Note"#8"留意"#9"NOTE"#10" NOTE"

R-lens tokens · positive

#1" note"#2" Note"#3" noting"#4" NOTE"#5"Note"#6" noted"#7" notes"#8"(note"#9" caution"#10"留意"

Skewness

+7.44

Excess kurtosis

229.80rank #565

Logcosh deviation

0.0949rank #993

Top-score occurrences · positive

  1. " note" of scripts↵↵Please note that, I'm↕ +48.00 (#1) · ⚡ 31.2%
  2. " Note"hi’s FAQ. Note that there are restrictions↕ +46.75 (#1) · ⚡ 43.9%
  3. " Note"0 in return. Note that many shops and↕ +46.75 (#1) · ⚡ 41.0%
  4. " note" easy updates. Please note that this is a↕ +46.25 (#1) · ⚡ 36.1%
  5. "Note" children too.↵Note however, that if↕ +44.75 (#1) · ⚡ 45.7%
  6. " note" Rail Station. Please note that CCMIT↕ +44.50 (#1) · ⚡ 38.5%
  7. " Note" calming the body. Note: Pregnant women↕ +43.75 (#1) · ⚡ 32.2%
  8. " note".↵↵*Please note that your prizes and↕ +43.50 (#1) · ⚡ 31.4%
  9. " note"!↵↵*Please note that while international readers↕ +43.00 (#1) · ⚡ 29.9%
  10. " Note"HandlingSettings argument. Note that the arguments are↕ +43.00 (#1) · ⚡ 32.8%
  11. " Note" 25. Note that we weren’t↕ +42.00 (#1) · ⚡ 33.4%
  12. " note" card.↵↵Please note that children above ↕ +42.00 (#1) · ⚡ 35.7%
  13. "Note"_HOST_URL. (Note that you may also↕ +41.75 (#1) · ⚡ 34.0%
  14. " mind" user (keep in mind spirit is highly rare↕ +41.50 (#1) · ⚡ 21.5%
  15. " mind".↵↵Keep in mind that I found out↕ +41.50 (#1) · ⚡ 22.2%
  16. " Note" individuals with disabilities." Note the word "qualified↕ +41.25 (#1) · ⚡ 33.8%
  17. " Note""," *"," * Note: native methods in↕ +40.50 (#1) · ⚡ 27.1%
  18. " noting"↵↵↵↵It is worth noting, meanwhile, that↕ +40.50 (#1) · ⚡ 24.6%
  19. "Note" finishing touches. (Note: the Grand Prix↕ +40.50 (#1) · ⚡ 28.1%
  20. "NOTE" BlackRock. (NOTE – Blackrock Secret↕ +40.50 (#1) · ⚡ 27.7%
Qwen 3.5 9B Base · L6/C2199

Counterfactual / hypothetical conditional

Confidence 5/5
Component profile — C2199 · layer 6 · positive tail

Logit-lens tokens · positive

#1"就好了"#2" hypoth"#3"的话"#4"ancier"#5" hätte"#6" could"#7"坐"#8" būtų"#9"hyp"#10"..."

R-lens tokens · positive

#1" would"#2" wouldn"#3" could"#4" couldn"#5" Would"#6"would"#7" really"#8" had"#9" weren"#10" hadn"

Skewness

+4.88

Excess kurtosis

160.67rank #1054

Logcosh deviation

0.0695rank #2200

Top-score occurrences · positive

  1. " so" the app wasn't so good, I would↕ +62.75 (#1) · ⚡ 25.6%
  2. " not" fighting↵↵“Had not nine of our nationals↕ +58.25 (#1) · ⚡ 26.4%
  3. " wasn" championship Chase if it wasn't for a small↕ +50.25 (#1) · ⚡ 26.6%
  4. " not" that, were it not for the requirement of↕ +48.00 (#1) · ⚡ 27.3%
  5. " it"."↵↵"Had it been left untreated,↕ +46.25 (#1) · ⚡ 23.1%
  6. " didn"ajib's dad) didn't invite and accept↕ +44.75 (#1) · ⚡ 27.3%
  7. " hadn" have happened if Jobs hadn’t misplaced his trust↕ +44.50 (#1) · ⚡ 24.7%
  8. " hadn")↵↵“If it hadn’t been for Chuck↕ +42.75 (#1) · ⚡ 21.8%
  9. " for"If it weren’t for the human power to↕ +42.25 (#1) · ⚡ 24.2%
  10. " were"œIf I were explaining sex to an↕ +41.00 (#1) · ⚡ 21.8%
  11. " been"↵↵"Had it been left untreated, Ahmed↕ +40.50 (#1) · ⚡ 23.8%
  12. " for" if it wasn't for Harper I would have↕ +40.50 (#1) · ⚡ 21.6%
  13. " not" immigrants had Schaaf not tipped off Oakland residents↕ +39.50 (#1) · ⚡ 22.1%
  14. " for" it hadn’t been for ECW, there↕ +39.50 (#1) · ⚡ 19.5%
  15. " been" if it hadn’t been for ECW,↕ +39.25 (#1) · ⚡ 24.5%
  16. " had", that if it had been made by any↕ +38.50 (#1) · ⚡ 25.8%
  17. " not"And, were it not for the width of↕ +38.25 (#1) · ⚡ 16.6%
  18. " were" does! If I were to phrase this as↕ +38.00 (#1) · ⚡ 22.1%
  19. " had" “omissions” had been discussed, several↕ +37.75 (#1) · ⚡ 25.8%
  20. " had", “If I had a son, he↕ +37.75 (#1) · ⚡ 23.5%
Qwen 3.5 9B Base · L7/C2650

/*

Confidence 5/5
Component profile — C2650 · layer 7 · negative tail

Logit-lens tokens · negative

#1"elho"#2"一章"#3"涯"#4"вори"#5"ection"#6"人来说"#7"een"#8"словно"#9" former"#10"ein"

R-lens tokens · negative

#1"eo"#2"een"#3" */"#4"ة"#5"eur"#6"o"#7"ascript"#8"eus"#9"uos"#10"ection"

Skewness

-6.81

Excess kurtosis

376.63rank #144

Logcosh deviation

0.0552rank #2651

Top-score occurrences · negative

  1. "/*"<|endoftext|>/*↵ * This library↕ -60.25 (#1) · ⚡ 56.6%
  2. "/*"<|endoftext|>/*↵ * Copyright (↕ -60.25 (#1) · ⚡ 56.6%
  3. "/*"<|endoftext|>/*↵Package logrus↕ -60.25 (#1) · ⚡ 56.6%
  4. "/*"<|endoftext|>/*↵* This file↕ -60.25 (#1) · ⚡ 56.6%
  5. "/*"<|endoftext|>/*--------------------------------*- C++↕ -60.25 (#1) · ⚡ 56.6%
  6. "/*"<|endoftext|>/*↵ * Licensed to↕ -60.25 (#1) · ⚡ 56.6%
  7. "/*"<|endoftext|>/*↵ * jcco↕ -60.25 (#1) · ⚡ 56.6%
  8. "/*"<|endoftext|>/*↵ * Copyright ↕ -60.25 (#1) · ⚡ 56.6%
  9. "/*"<|endoftext|>/*↵ * Copyright (↕ -60.25 (#1) · ⚡ 56.6%
  10. "/*"<|endoftext|>/*↵ * Licensed to↕ -60.25 (#1) · ⚡ 56.6%
  11. "/*"<|endoftext|>/*global doSweep↕ -60.25 (#1) · ⚡ 56.6%
  12. "/*"<|endoftext|>/*↵Copyright (c↕ -60.25 (#1) · ⚡ 56.6%
  13. "/*"<|endoftext|>/*↵ * Copyright (↕ -60.25 (#1) · ⚡ 56.6%
  14. "/*"<|endoftext|>/*↵ * The Clear↕ -60.25 (#1) · ⚡ 56.6%
  15. "/*"<|endoftext|>/*↵ * Copyright (↕ -60.25 (#1) · ⚡ 56.6%
  16. "/*"<|endoftext|>/* -*- Mode: C↕ -60.25 (#1) · ⚡ 56.6%
  17. "/*"<|endoftext|>/*#######################################################↵ *↕ -60.25 (#1) · ⚡ 56.6%
  18. "/*"<|endoftext|>/*↵Copyright 2↕ -60.25 (#1) · ⚡ 56.6%
  19. "/*"<|endoftext|>/* eslint-disable eslint-comments↕ -60.25 (#1) · ⚡ 56.6%
  20. "/*"<|endoftext|>/*↵Copyright 2↕ -60.25 (#1) · ⚡ 56.6%
Qwen 3.5 9B Base · L7/C3045

Helicopter / rotorcraft

Confidence 5/5
Component profile — C3045 · layer 7 · positive tail

Logit-lens tokens · positive

#1"吊"#2"estro"#3"ansi"#4"gade"#5"uhr"#6"っこ"#7"dep"#8"ports"#9"anmar"#10"ensem"

R-lens tokens · positive

#1"Rotor"#2" rotor"#3"ยนต์"#4"裳"#5" skies"#6"空中"#7" loft"#8" лопа"#9"bird"#10"直升机"

Skewness

+2.31

Excess kurtosis

77.51rank #1811

Logcosh deviation

0.0460rank #3046

Top-score occurrences · positive

  1. "ip"ports," aka helipads. The company↕ +49.00 (#1) · ⚡ 20.1%
  2. " helicopter" invention relates to a helicopter rotor of the bearing↕ +49.00 (#1) · ⚡ 19.8%
  3. " cyclic" speed and undergoing commanded cyclic pitch changes as well↕ +47.50 (#1) · ⚡ 17.3%
  4. "off"OL (vertical takeoff and landing) designs↕ +43.00 (#1) · ⚡ 13.1%
  5. "rot" bits. Quad-rotor designs like Bell↕ +43.00 (#1) · ⚡ 15.5%
  6. " rotor" formation of ice on rotor blades and other surfaces↕ +41.75 (#1) · ⚡ 25.4%
  7. " rotor", and heavier main rotor blade undergoes significant↕ +41.50 (#1) · ⚡ 24.4%
  8. "apping" of the blade flapping wherein relative bending takes↕ +40.50 (#1) · ⚡ 16.4%
  9. " helicopters" 27% helicopters.↵↵History↵↕ +40.25 (#1) · ⚡ 16.9%
  10. "Hel" every day.↵↵Helicopter Tours. Ever↕ +40.00 (#1) · ⚡ 20.1%
  11. " One" dog to the Marine One chopper in ↕ +39.25 (#1) · ⚡ 7.5%
  12. " hel"Skyports," aka helipads. The↕ +36.50 (#1) · ⚡ 21.5%
  13. "2" Bell at CES 2019 in↕ +36.50 (#1) · ⚡ 18.8%
  14. " head"" of that rotor head is depicted as FIG↕ +35.50 (#1) · ⚡ 17.2%
  15. " rotor" suitable attachment to the rotor hub and rotor drive↕ +35.00 (#1) · ⚡ 22.4%
  16. " taxis" a fleet of air taxis in operation by ↕ +31.75 (#1) · ⚡ 10.4%
  17. " rotor" pertaining to hingeless rotor blade attachments and controls↕ +30.88 (#1) · ⚡ 17.4%
  18. " rotor"jet engines on the rotor tips for powered take↕ +30.75 (#1) · ⚡ 17.9%
  19. " Bell"rotor designs like Bell's, much less↕ +30.75 (#1) · ⚡ 12.9%
  20. " rotor" The antitorque rotor represents a sizeable↕ +30.25 (#1) · ⚡ 18.2%
Qwen 3.5 9B Base · L7/C3826

?

Confidence 0/5
Component profile — C3826 · layer 7 · negative tail

Logit-lens tokens · negative

#1".getClass"#2"也不行"#3"勉强"#4"requ"#5"�"#6"sic"#7"让它们"#8"陈小"#9"jabi"#10"vir"

R-lens tokens · negative

#1"ehand"#2"togroup"#3"��"#4"]\")."#5"потре"#6"\f\n"#7"].'"#8"/DTD"#9"]\")"#10"��"

Skewness

-0.02

Excess kurtosis

1.74rank #3981

Logcosh deviation

0.0199rank #3827

Top-score occurrences · negative

  1. " less" no affection, much less sex — and he↕ -10.62 (#5) · ⚡ 1.1%
  2. " hey" be fun, but hey, it really was↕ -9.75 (#7) · ⚡ 1.2%
  3. " Well" could have been. Well, duh!↕ -8.94 (#4) · ⚡ 0.8%
  4. " less" any kind, much less a sports bra,↕ -8.88 (#10) · ⚡ 0.8%
  5. " certainly" for my taste - certainly not the trend-driven↕ -8.69 (#7) · ⚡ 1.0%
  6. " guess" better, but I guess you can't have↕ -8.62 (#10) · ⚡ 0.9%
  7. " out" and as it turned out, she wasn't↕ -8.06 (#10) · ⚡ 0.7%
  8. " get" continually play. I get that they want to↕ -7.97 (#7) · ⚡ 1.3%
  9. " not" makes work difficult if not impossible, the high↕ -7.94 (#8) · ⚡ 0.8%
  10. " for"’ll cave. Good for Heboric and↕ -7.91 (#7) · ⚡ 0.8%
  11. " long", but for how long? I get the↕ -7.84 (#7) · ⚡ 0.9%
  12. " mentioning" power, not even mentioning other difficulties, like↕ -7.81 (#11) · ⚡ 0.5%
  13. " guess" a novel. I guess that's something↵↕ -7.81 (#9) · ⚡ 0.9%
  14. " guess" was leaving. I guess I won't care↕ -7.62 (#9) · ⚡ 1.0%
  15. " чём" не работают, в чём может быть проблема ?↕ -7.50 (#10) · ⚡ 0.5%
  16. " could" and nightlife. He could hear something rattling↕ -7.38 (#6) · ⚡ 1.4%
  17. " get" to capture..you get anywhere near these birds↕ -7.34 (#6) · ⚡ 1.1%
  18. " чём" заменить картину. В чём может быть проблема в↕ -7.19 (#15) · ⚡ 0.4%
  19. " not" are comprised largely if not exclusively of PrP↕ -7.12 (#15) · ⚡ 0.4%
  20. " little" flames. Maybe a little orange glow when she↕ -7.09 (#6) · ⚡ 0.9%
Qwen 3.5 9B Base · L8/C935

Repeated structured code / machine output

Confidence 4/5
Component profile — C935 · layer 8 · positive tail

Logit-lens tokens · positive

#1"苦辣"#2"InSeconds"#3"ersions"#4"ugins"#5"เช่นกัน"#6"至关"#7"øk"#8"��"#9"---</"#10"ุกร"

R-lens tokens · positive

#1" specifie"#2"Agen"#3" Ress"#4"��"#5"<quote"#6"üyada"#7" videok"#8"yms"#9"стребо"#10"Provin"

Skewness

+3.07

Excess kurtosis

27.33rank #2938

Logcosh deviation

0.0886rank #936

Top-score occurrences · positive

  1. " SET" quest_template_addon SET RequiredSkillID =↕ +19.38 (#1) · ⚡ 11.0%
  2. " SET" quest_template_addon SET RequiredSkillID =↕ +19.00 (#1) · ⚡ 10.2%
  3. " database"Data↵!!!Unknown database.↵Alias:↕ +18.62 (#1) · ⚡ 16.6%
  4. " database"Data↵!!!Unknown database.↵Alias:↕ +18.62 (#1) · ⚡ 16.1%
  5. " database"Data↵!!!Unknown database.↵Alias:↕ +18.50 (#1) · ⚡ 16.1%
  6. " database"Data↵!!!Unknown database.↵Alias:↕ +18.25 (#1) · ⚡ 15.7%
  7. " SET" quest_template_addon SET RequiredSkillID =↕ +18.12 (#1) · ⚡ 9.2%
  8. " database"Data↵!!!Unknown database.↵Alias:↕ +18.00 (#1) · ⚡ 15.4%
  9. "Alias"Unknown database.↵Alias: dbCaps9↕ +17.88 (#1) · ⚡ 15.4%
  10. " database"Data↵!!!Unknown database.↵Alias:↕ +17.88 (#1) · ⚡ 15.5%
  11. "ID"on SET RequiredSkillID = 25↕ +17.88 (#1) · ⚡ 15.8%
  12. "on"UPDATE quest_template_addon SET RequiredSkillID↕ +17.62 (#1) · ⚡ 16.1%
  13. "Alias"Unknown database.↵Alias: dbCaps9↕ +17.62 (#1) · ⚡ 14.5%
  14. "Alias"Unknown database.↵Alias: dbCaps9↕ +17.50 (#1) · ⚡ 14.3%
  15. "read">↵ <read><delim>\↕ +17.38 (#1) · ⚡ 6.1%
  16. "read">↵ <read><delim>\↕ +17.38 (#1) · ⚡ 7.1%
  17. " database"Data↵!!!Unknown database.↵Alias:↕ +17.12 (#1) · ⚡ 13.5%
  18. "ID"on SET RequiredSkillID = 25↕ +17.00 (#1) · ⚡ 14.3%
  19. "Path" "corpusPath": "/core/↕ +17.00 (#1) · ⚡ 12.0%
  20. "ID"on SET RequiredSkillID = 25↕ +16.88 (#1) · ⚡ 13.0%
Qwen 3.5 9B Base · L9/C2175

Children’s / preschool educational content

Confidence 4/5
Component profile — C2175 · layer 9 · negative tail

Logit-lens tokens · negative

#1"绘本"#2"创设"#3"STEM"#4"适龄"#5"爸爸妈妈"#6"引导幼儿"#7" preschool"#8"也很好"#9"阿联"#10"adelphia"

R-lens tokens · negative

#1" preschool"#2" kids"#3" Kindergarten"#4"绘本"#5" kindergarten"#6"对幼儿"#7"学龄"#8"kids"#9" children"#10" дошко"

Skewness

-2.68

Excess kurtosis

58.13rank #1772

Logcosh deviation

0.0598rank #2176

Top-score occurrences · negative

  1. "no" lawn.↵↵Anno's Math Games↵↕ -51.75 (#1) · ⚡ 13.1%
  2. "?" Stars in the Sky?↵Multicultural↵↕ -39.50 (#1) · ⚡ 18.5%
  3. " book".↵↵This colourful book provides children with a↕ -38.75 (#1) · ⚡ 17.2%
  4. "\n\n" Simms Taback↵↵Reviewed by Colin C↕ -37.00 (#1) · ⚡ 18.8%
  5. "p"7 (32p) also paper↵↕ -35.50 (#1) · ⚡ 19.2%
  6. " books" the use of picture books for preschoolers through↕ -34.75 (#1) · ⚡ 16.3%
  7. " ABC"Dr Seuss’s ABC: AR Version↵↵↕ -33.50 (#1) · ⚡ 8.5%
  8. " a" book provides children with a basic understanding of the↕ -33.25 (#1) · ⚡ 24.6%
  9. "." night to count stars.↵↵17 Kings↕ -32.75 (#1) · ⚡ 19.3%
  10. "bridge"ildes↵Charlesbridge 199↕ -32.75 (#2) · ⚡ 7.3%
  11. " boy"p)↵A boy and his father who↕ -32.50 (#1) · ⚡ 16.7%
  12. " book" Rights era), the book brings to the surface↕ -31.75 (#1) · ⚡ 17.4%
  13. "icultural" Sky?↵Multicultural↵by Lenny↕ -31.00 (#1) · ⚡ 10.5%
  14. ")" (63 p)↵Description of the↕ -30.38 (#1) · ⚡ 17.7%
  15. " Anno"by Mitsumasa Anno↵Philomel↕ -30.25 (#1) · ⚡ 5.2%
  16. " counts" gets seven when he counts the toes on one↕ -29.62 (#1) · ⚡ 7.7%
  17. "1"Scholastic 1995 (↕ -29.38 (#1) · ⚡ 22.3%
  18. " Hat" “Cat in the Hat” narrated on↕ -29.12 (#2) · ⚡ 2.8%
  19. " the" patted it on the head.↵↵My↕ -28.25 (#1) · ⚡ 18.3%
  20. " for" each picture to stand for more than 1↕ -28.25 (#1) · ⚡ 6.0%
Qwen 3.5 9B Base · L10/C170

Footnote / reference marker

Confidence 5/5
Component profile — C170 · layer 10 · positive tail

Logit-lens tokens · positive

#1" ,["#2"isal"#3" Equipe"#4"昕"#5"foot"#6"anja"#7"verzeichnis"#8"青松"#9"imba"#10" footnote"

R-lens tokens · positive

#1" footnote"#2"Foot"#3"foot"#4" disclaimer"#5"-note"#6"§"#7"_foot"#8"注明"#9" Foot"#10"“("

Skewness

+20.95

Excess kurtosis

1462.90rank #13

Logcosh deviation

0.1221rank #171

Top-score occurrences · positive

  1. "^"12pj][^2]. Explo↕ +105.00 (#1) · ⚡ 53.1%
  2. "1" filed a reply[1] to plaintiffs'↕ +100.50 (#1) · ⚡ 66.4%
  3. "[^" potential at finite temperature[^1] [@K↕ +99.00 (#1) · ⚡ 53.9%
  4. "[^"aseous Molecules[^1]. For each↕ +96.50 (#1) · ⚡ 52.0%
  5. "1" to employees. 1↵ In December↕ +89.00 (#1) · ⚡ 62.5%
  6. "[^" the univariate case[^4]. The approach↕ +88.00 (#1) · ⚡ 54.7%
  7. "1" Commission's order.[1]↵On December↕ +86.50 (#1) · ⚡ 60.9%
  8. "1" the ensuing hearing.[1] Faraghar↕ +85.50 (#1) · ⚡ 63.3%
  9. "1"214.[1] On October ↕ +83.00 (#1) · ⚡ 55.9%
  10. "1"006),1 and 28↕ +81.00 (#1) · ⚡ 51.2%
  11. "^" Limit Theorem [^1] '↵↕ +81.00 (#1) · ⚡ 28.3%
  12. "1" COMPTON,[1] LACY,↕ +77.00 (#1) · ⚡ 52.3%
  13. "1" (TMC).[1] St. Joseph↕ +75.50 (#1) · ⚡ 59.8%
  14. "1".16.[1] When the order↕ +75.00 (#1) · ⚡ 55.9%
  15. "1" of liability, [1] something necessary both↕ +74.50 (#1) · ⚡ 55.1%
  16. "1"-law clause. 1↵ Next,↕ +73.00 (#1) · ⚡ 55.1%
  17. "1" in this case.1↵↵I. ↕ +72.50 (#1) · ⚡ 51.6%
  18. "[^" Pierre E. Jacob[^1]\↵ ↕ +69.50 (#1) · ⚡ 23.9%
  19. "^" (SRM)[^14^](↕ +67.50 (#1) · ⚡ 33.2%
  20. "2" the shopping center.2 It contained a↕ +67.50 (#1) · ⚡ 54.7%
Qwen 3.5 9B Base · L10/C4002

?

Confidence 0/5
Component profile — C4002 · layer 10 · positive tail

Logit-lens tokens · positive

#1"REFERRED"#2"IGA"#3"zysz"#4"imba"#5"ässe"#6"因为在"#7"想说"#8"众筹"#9"inne"#10"珍"

R-lens tokens · positive

#1"-esque"#2"ABCDEFG"#3" Inclus"#4"Whilst"#5" Certif"#6"名為"#7" Distribu"#8"-ish"#9"ún"#10"[…]"

Skewness

+0.03

Excess kurtosis

1.38rank #4000

Logcosh deviation

0.0146rank #4003

Top-score occurrences · positive

  1. "J"16 The MathJax Consortium↵ *↕ +16.38 (#8) · ⚡ 0.4%
  2. "ovan": Tanya Khovanova writes More and↕ +9.62 (#12) · ⚡ 0.3%
  3. "ette" part of the Nette Framework (https://↕ +9.38 (#27) · ⚡ 0.2%
  4. " Dorm" La Belle au Bois Dormant, Jeannot↕ +8.69 (#23) · ⚡ 0.3%
  5. "oms"==========↵↵Diatoms are important primary producers↕ +8.19 (#28) · ⚡ 0.2%
  6. " at" and while you're at it, check out↕ +7.97 (#8) · ⚡ 0.5%
  7. "us" extradigital glomus tumor and proceed to↕ +7.97 (#12) · ⚡ 0.3%
  8. " uptake"obutyric acid uptake in the dorsomed↕ +7.69 (#10) · ⚡ 0.4%
  9. "cone" through inhibition of gluconeogenesis \[[@↕ +7.59 (#16) · ⚡ 0.3%
  10. "ioms"lex Dictionary of Idioms. © 2↕ +7.59 (#33) · ⚡ 0.2%
  11. "9" Form 1099s will become↕ +7.44 (#25) · ⚡ 0.2%
  12. "ube"class acl extends rcube_plugin↵{↵↕ +7.38 (#157) · ⚡ 0.1%
  13. "J".cell_output .MathJax_Display {↵↕ +7.34 (#31) · ⚡ 0.2%
  14. " sensory", and thermal quantitative sensory testing. On two↕ +7.16 (#8) · ⚡ 0.5%
  15. "fmt"-openapi/strfmt"↵ "github↕ +7.00 (#32) · ⚡ 0.2%
  16. " ':" strpos($userInfo, ':'))) {↵ ↕ +6.97 (#26) · ⚡ 0.2%
  17. "otech" al., Nature Biotech. 25↕ +6.94 (#190) · ⚡ 0.1%
  18. "Prefix" IconDefinition, IconPrefix, IconName }↕ +6.91 (#48) · ⚡ 0.2%
  19. " vestib" a unilateral loss of vestibular function. H↕ +6.88 (#14) · ⚡ 0.3%
  20. "b"support.microsoft.com/kb/167↕ +6.84 (#19) · ⚡ 0.3%
Qwen 3.5 9B Base · L13/C637

Online forum / discussion thread

Confidence 4/5
Component profile — C637 · layer 13 · negative tail

Logit-lens tokens · negative

#1" Replies"#2"回复"#3" Forums"#4"/forum"#5" Moderator"#6"回復"#7" reply"#8" moderator"#9" replies"#10"回覆"

R-lens tokens · negative

#1" Forums"#2" alot"#3" kinda"#4" forum"#5"回复"#6"/forum"#7"forums"#8" REALLY"#9" forums"#10" really"

Skewness

-4.52

Excess kurtosis

87.23rank #606

Logcosh deviation

0.0806rank #638

Top-score occurrences · negative

  1. "\n\n" in mouth brigade?↵↵I do not cond↕ -40.00 (#1) · ⚡ 22.9%
  2. "\n\n" along those lines.↵↵Spoiler:↵↵Looks↕ -38.25 (#1) · ⚡ 20.5%
  3. "\n\n" ausgespielt.↵↵Some commenters seem↕ -34.50 (#1) · ⚡ 13.6%
  4. "\n\n" ever dream of.↵↵I was just being↕ -33.50 (#1) · ⚡ 19.0%
  5. "\n\n" it was accurate.↵↵Quote:↵↵As↕ -33.00 (#1) · ⚡ 21.7%
  6. " sites"↵↵Share on other sites↵↵She has famously↕ -32.75 (#2) · ⚡ 4.6%
  7. "\n\n" one over it?↵↵Click to expand...↕ -32.75 (#1) · ⚡ 12.2%
  8. "\n\n" the whole album.↵↵It amazes me↕ -32.50 (#1) · ⚡ 22.5%
  9. "\n\n" curcles more*↵↵_________________Shattered↕ -32.00 (#1) · ⚡ 14.8%
  10. "\n\n" doing for years.↵↵Do you issue actual↕ -31.25 (#1) · ⚡ 14.3%
  11. "\n\n" is too open.↵↵"When you hear my↕ -31.25 (#1) · ⚡ 17.4%
  12. " sites"↵↵Share on other sites↵↵They had to↕ -31.00 (#1) · ⚡ 11.4%
  13. "\n\n" explaining the context.↵↵It mainly looks '↕ -30.75 (#1) · ⚡ 16.5%
  14. "\n\n" we need here.↵↵This isn't the↕ -30.62 (#1) · ⚡ 22.4%
  15. "\n\n" ready.↵↵Comment↵↵I have the same↕ -30.12 (#2) · ⚡ 10.8%
  16. "\n\n" tractor as Ford.↵↵My brother runs a↕ -30.00 (#1) · ⚡ 14.3%
  17. "\n\n". Ha ha.↵↵I finished season ↕ -29.88 (#1) · ⚡ 17.2%
  18. "well" opinions biased.↵↵well, to be fair↕ -29.88 (#1) · ⚡ 12.7%
  19. " sites"↵↵Share on other sites↵↵I don't↕ -29.62 (#2) · ⚡ 10.0%
  20. "\n\n" it was accurate.↵↵Oh, I thought↕ -29.50 (#1) · ⚡ 14.9%
Qwen 3.5 9B Base · L13/C889

Commented-out code

Confidence 4/5
Component profile — C889 · layer 13 · positive tail

Logit-lens tokens · positive

#1" Uncomment"#2" uncomment"#3"kees"#4"FIXME"#5" TODO"#6"TODO"#7"iou"#8" jailed"#9"contri"#10"andbox"

R-lens tokens · positive

#1" TODO"#2" Uncomment"#3" uncomment"#4"TODO"#5"*/"#6"->"#7"////"#8" FIXME"#9" -->"#10"\t"

Skewness

+5.52

Excess kurtosis

175.38rank #283

Logcosh deviation

0.0716rank #890

Top-score occurrences · positive

  1. "basic" { ↵ //basicEffect.VertexColorEnabled↕ +65.50 (#1) · ⚡ 47.3%
  2. "Log" {↵ //Log.e(debugTag,↕ +52.25 (#1) · ⚡ 29.3%
  3. " $('." });↵ // $('.wrapper').delay()↕ +49.50 (#1) · ⚡ 39.3%
  4. " "];↵ // CHECK IF FILE EXISTS↕ +47.50 (#1) · ⚡ 24.6%
  5. ";"use_attributes=0;↵ //use↕ +46.25 (#1) · ⚡ 34.8%
  6. " auto"14↵ // auto counter = [&i↕ +44.00 (#1) · ⚡ 27.3%
  7. "params" {};↵ //params.filePath = "files↕ +44.00 (#1) · ⚡ 29.9%
  8. " gem" runtimes↵# gem 'therubyr↕ +43.50 (#1) · ⚡ 23.8%
  9. " <">↵ <!-- <span id="click↕ +43.25 (#1) · ⚡ 26.2%
  10. "Process".start(); ↵//Process pr = rt.exec↕ +42.25 (#1) · ⚡ 31.2%
  11. " ="↵ //entries = Directory.GetFiles(sourceDirectory↕ +41.50 (#1) · ⚡ 29.3%
  12. "include"_H↵↵// #include "nrf.h↕ +41.25 (#1) · ⚡ 20.6%
  13. " if"↵ /*↵ if(data[amonst↕ +40.75 (#1) · ⚡ 29.3%
  14. "read" shift↵ #readlink -f↵↕ +40.50 (#1) · ⚡ 23.8%
  15. ";"Right, driveLeft;↵↵ boolean has↕ +40.50 (#1) · ⚡ 25.4%
  16. "();" //Learn_to_use();↵↵ break;↕ +39.25 (#1) · ⚡ 21.0%
  17. ");">(100);↵ real_type↕ +39.00 (#1) · ⚡ 25.8%
  18. ">" //List<String> bins_bits = new↕ +38.50 (#1) · ⚡ 32.4%
  19. "Curve" //drawCurve(new c.Line([↕ +38.00 (#1) · ⚡ 23.7%
  20. "=" //use_attributes=0;↵ ↕ +37.75 (#1) · ⚡ 24.9%
Qwen 3.5 9B Base · L15/C1201

Enumerating numbers

Confidence 1/5
Component profile — C1201 · layer 15 · positive tail

Logit-lens tokens · positive

#1"乃至"#2"以此"#3"一直到"#4"直至"#5"再到"#6"依"#7"等不同"#8"以至"#9" etc"#10"而至"

R-lens tokens · positive

#1" etc"#2"乃至"#3" onwards"#4"一直到"#5" intermediate"#6"等不同"#7" seterusnya"#8" onward"#9"直至"#10"再到"

Skewness

+2.25

Excess kurtosis

27.98rank #1452

Logcosh deviation

0.0596rank #1202

Top-score occurrences · positive

  1. "B" in ('A','B','C','D↕ +26.88 (#1) · ⚡ 9.5%
  2. "1"0], data[1], data[2↕ +25.38 (#1) · ⚡ 8.3%
  3. "2"get_rating_count(2);↵$rating↕ +25.25 (#1) · ⚡ 14.1%
  4. "0" mg, 10 mg, 2↕ +24.12 (#1) · ⚡ 11.6%
  5. "5" Bob Lindsay↵↵65-69 (↕ +23.12 (#1) · ⚡ 5.8%
  6. "0"5, 50, 10↕ +22.75 (#1) · ⚡ 11.7%
  7. "2" (1, 2, 3,↕ +22.50 (#1) · ⚡ 10.1%
  8. "3"get_rating_count(3);↵$rating↕ +22.12 (#1) · ⚡ 10.1%
  9. "2"1 d | D2 d | D3↕ +21.62 (#1) · ⚡ 6.3%
  10. "2" col1, col2, col3,↕ +21.50 (#1) · ⚡ 9.8%
  11. " minutes" units of seconds, minutes, and hours.↕ +21.50 (#1) · ⚡ 8.7%
  12. "8"4, 48, and 7↕ +21.25 (#1) · ⚡ 8.2%
  13. "2"1], data[2], data[3↕ +21.25 (#1) · ⚡ 6.5%
  14. "2"1 _002 if they exist,↕ +21.12 (#1) · ⚡ 6.6%
  15. " moderate" scale of mild, moderate and severe. The↕ +21.00 (#1) · ⚡ 9.7%
  16. " =" 1, low = 2, normal↕ +21.00 (#1) · ⚡ 6.2%
  17. "r" $g$, $r$, $i$,↕ +20.88 (#1) · ⚡ 6.1%
  18. "3"i=2,3,4,5↕ +20.50 (#1) · ⚡ 9.1%
  19. "2" 1, 2, 3,↕ +20.50 (#1) · ⚡ 7.1%
  20. "0"5, 30 and 35↕ +20.38 (#1) · ⚡ 9.9%
Qwen 3.5 9B Base · L16/C61

Wikipedia category / category-page

Confidence 4/5
Component profile — C61 · layer 16 · positive tail

Logit-lens tokens · positive

#1" 이름을"#2"{{{"#3"{{"#4"tribusi"#5"この記事"#6"希望对大家"#7"IOC"#8"ことになります"#9" stub"#10"的朋友可以"

R-lens tokens · positive

#1" Uncategorized"#2",List"#3" Wikimedia"#4" freig"#5"herits"#6">List"#7" classific"#8"この記事"#9" LGBT"#10" 이름을"

Skewness

+10.41

Excess kurtosis

230.39rank #151

Logcosh deviation

0.1507rank #62

Top-score occurrences · positive

  1. "ums"↵Category:Museums in Saskatoon↵↕ +36.75 (#1) · ⚡ 25.4%
  2. "ers"↵Category:Painters from Suzhou↵↕ +36.25 (#1) · ⚡ 29.1%
  3. " institutions"Category:Educational institutions established in 1↕ +35.75 (#1) · ⚡ 27.5%
  4. "Companies" Russia↵Category:Companies based in Moscow↵↕ +34.75 (#1) · ⚡ 28.1%
  5. " structures":Buildings and structures in the Province of↕ +34.50 (#1) · ⚡ 20.6%
  6. "icians"↵Category:Politicians from Lancaster, Pennsylvania↕ +34.50 (#1) · ⚡ 27.7%
  7. " areas"↵Category:Protected areas established in 1↕ +34.25 (#1) · ⚡ 24.7%
  8. " leagues"↵Category:Sports leagues established in 2↕ +34.00 (#1) · ⚡ 22.1%
  9. "ets"↵Category:Poets Laureate of Louisiana↕ +33.50 (#1) · ⚡ 26.8%
  10. "ays"↵Category:Plays by Voltaire↵↕ +33.50 (#1) · ⚡ 24.8%
  11. " football"↵Category:Association football clubs established in ↕ +33.00 (#1) · ⚡ 22.0%
  12. " structures":Buildings and structures in San Luis Ob↕ +33.00 (#1) · ⚡ 19.4%
  13. "igious"↵Category:Religious buildings and structures in↕ +33.00 (#1) · ⚡ 25.4%
  14. "crap"Category:Skyscrapers in Troy,↕ +32.75 (#1) · ⚡ 20.7%
  15. " actors"↵Category:Male actors from Surrey↵Category↕ +32.75 (#1) · ⚡ 23.1%
  16. " agencies"↵Category:Government agencies of the Netherlands↕ +32.75 (#1) · ⚡ 26.6%
  17. "-century":21st-century women artists↵Category↕ +32.75 (#1) · ⚡ 18.4%
  18. " male"20th-century male singers↕ +32.50 (#1) · ⚡ 19.5%
  19. "Scientists"ians↵Category:Scientists from Kerala↕ +32.50 (#1) · ⚡ 27.5%
  20. " companies"Category:Publishing companies established in 1↕ +32.25 (#1) · ⚡ 30.1%
Qwen 3.5 9B Base · L16/C244

Repeated previously mentioned name entity

Confidence 4/5
Component profile — C244 · layer 16 · negative tail

Logit-lens tokens · negative

#1"FilterWhere"#2"opedia"#3"ier"#4"adel"#5"idão"#6"wassen"#7"uit"#8"**–"#9"oplast"#10"**-"

R-lens tokens · negative

#1"FilterWhere"#2"сполни"#3"\"urls"#4"$fdata"#5"BitFields"#6"<|box_end|>"#7"PropertyParams"#8"GuidId"#9" **「"#10"\"indices"

Skewness

-3.25

Excess kurtosis

22.96rank #1693

Logcosh deviation

0.1021rank #245

Top-score occurrences · negative

  1. " De" occurred and when the Deccan Traps↕ -15.00 (#1) · ⚡ 6.7%
  2. " cig" forecasting and planning for ciguatera outbreaks.”↕ -13.69 (#1) · ⚡ 5.2%
  3. " def" The success rate of defibrillation was ↕ -13.69 (#1) · ⚡ 4.2%
  4. " He" Y there is a Heegaard splitting such↕ -13.62 (#2) · ⚡ 3.3%
  5. " De" tipping point and the Decepticon army appeared↕ -13.62 (#1) · ⚡ 3.6%
  6. " pil" the treatment of limited pilonidal disease.↕ -13.31 (#1) · ⚡ 4.8%
  7. " meth"forms catalyzing methadone metabolism by↕ -13.25 (#1) · ⚡ 4.6%
  8. " do" from reduced intracellular doxorubicin accumulation↕ -13.12 (#1) · ⚡ 4.1%
  9. " Cal" for all semiample Calabi-Yau hypers↕ -13.06 (#1) · ⚡ 3.8%
  10. " Sum" frames.↵↵The Sumatran rhino↕ -13.00 (#1) · ⚡ 5.4%
  11. "ent". Because the↵entorhinal cortex↕ -12.94 (#2) · ⚡ 3.9%
  12. " Wel" Center include expanding the Welga! Project to↕ -12.88 (#1) · ⚡ 4.1%
  13. " Sch"function.↵The Schrodinger equation gives↕ -12.88 (#1) · ⚡ 3.0%
  14. " Is" the time, the Ismailis were somewhat↕ -12.81 (#1) · ⚡ 4.3%
  15. " Nat" was orally sentenced in Natrona County to↕ -12.75 (#3) · ⚡ 2.7%
  16. " cat" some assays (especially cathepsin L↕ -12.75 (#1) · ⚡ 3.6%
  17. " am" The MICs for amphotericin B↕ -12.69 (#1) · ⚡ 3.7%
  18. " Ant"ukahara↵↵Former Antiquorum officials,↕ -12.69 (#1) · ⚡ 4.4%
  19. " ev" Then the contribution of evanes↕ -12.62 (#1) · ⚡ 5.0%
  20. " str"0%, while the stratum corneum↕ -12.56 (#2) · ⚡ 2.0%
Qwen 3.5 9B Base · L16/C1365

Closing tag </

Confidence 4/5
Component profile — C1365 · layer 16 · negative tail

Logit-lens tokens · negative

#1"<|box_end|>"#2"uja"#3"vern"#4"说吧"#5" hidrat"#6"放矢"#7"니다"#8"�"#9"jder"#10"价差"

R-lens tokens · negative

#1"<|box_end|>"#2"_DIP"#3" sostanzi"#4"#__"#5"全心全意"#6"PostalCodes"#7".gdx"#8"ICODE"#9"的"#10"сполни"

Skewness

-2.80

Excess kurtosis

45.07rank #925

Logcosh deviation

0.0563rank #1366

Top-score occurrences · negative

  1. "</"abbr> Patent Policy</a>.↵ ↕ -30.00 (#1) · ⚡ 21.5%
  2. "></"98</pre></p>↵ ↕ -28.50 (#1) · ⚡ 17.4%
  3. "</"ed Zepplin</artist>↵ ↕ -27.75 (#1) · ⚡ 22.5%
  4. ".</" change contrast and brightness.</header>↵ ↕ -27.12 (#1) · ⚡ 13.4%
  5. "\"></"" placeholder="请输入"></input>↵ ↕ -27.00 (#1) · ⚡ 14.1%
  6. "</">multipart/form-data</pre></h2↕ -26.88 (#1) · ⚡ 12.7%
  7. "</" 12 )</title>↵<meta↕ -26.50 (#1) · ⚡ 15.2%
  8. ")</"5.0 API)</title>↵<meta↕ -26.50 (#1) · ⚡ 16.7%
  9. "</".4.0</version>↵</↕ -25.75 (#1) · ⚡ 16.9%
  10. "</"U.S. Dollars</ReportingCurrency>↵↕ -25.62 (#1) · ⚡ 18.5%
  11. "</"_page_menu ">Primary</a>↵ ↕ -25.50 (#1) · ⚡ 11.5%
  12. "</" ?>">share on facebook</a>↵↵Link↕ -25.25 (#1) · ⚡ 14.6%
  13. "</">\x0a</delim><match↕ -25.25 (#1) · ⚡ 19.7%
  14. "</"="">Hearthstone</label>↵ ↕ -25.12 (#1) · ⚡ 12.2%
  15. "</"%">Something goes wrong</div>↵ ↕ -24.62 (#1) · ⚡ 9.8%
  16. "}</"LANG.cartdomainsconfig}</h1>↵↕ -24.62 (#1) · ⚡ 11.5%
  17. "</"7098</pre></p>↕ -24.62 (#1) · ⚡ 10.3%
  18. "</">Mothership</name>↵ ↕ -24.62 (#1) · ⚡ 11.0%
  19. "</" JDK 12</strong> </div↕ -24.38 (#1) · ⚡ 15.0%
  20. "</"alibaba Directory Reference</div> </↕ -24.38 (#1) · ⚡ 11.5%
Qwen 3.5 9B Base · L16/C3363

?

Confidence 0/5
Component profile — C3363 · layer 16 · negative tail

Logit-lens tokens · negative

#1"ongan"#2" lately"#3"áng"#4"бол"#5"ubr"#6"最新章节"#7" interim"#8"ubb"#9"armacy"#10"weite"

R-lens tokens · negative

#1" preparations"#2"准备的"#3"的准备"#4" preparation"#5" awaiting"#6"来临"#7"備"#8" preparing"#9" upcoming"#10"准备好的"

Skewness

-0.00

Excess kurtosis

0.85rank #3550

Logcosh deviation

0.0117rank #3364

Top-score occurrences · negative

  1. " dept"es-what? dept↵↵The Los Angeles↕ -8.19 (#59) · ⚡ 0.2%
  2. " bond" he was able to bond out. But before↕ -7.59 (#4) · ⚡ 0.8%
  3. " out" was able to bond out. But before he↕ -6.84 (#4) · ⚡ 0.8%
  4. "enary", marking the centenary year of Aston Martin↕ -6.62 (#4) · ⚡ 0.7%
  5. "home"ollah Ruhollah Khomeini afterward. After↕ -6.41 (#9) · ⚡ 0.6%
  6. " block" are on the chopping block in $15↕ -6.12 (#18) · ⚡ 0.4%
  7. " of"^th^-week of age, female age↕ -5.91 (#5) · ⚡ 0.6%
  8. " distract" him to attempt to distract the public from the↕ -5.84 (#7) · ⚡ 0.6%
  9. " suspect" if you have or suspect you might have a↕ -5.84 (#18) · ⚡ 0.4%
  10. " innocent" regularly lynched an innocent man throughout history.↕ -5.81 (#6) · ⚡ 0.8%
  11. " rel" as diagnosis of molecular relapse now preceded the↕ -5.66 (#6) · ⚡ 0.6%
  12. " rise" is always an initial rise in numbers,” said↕ -5.62 (#11) · ⚡ 0.5%
  13. "J"16 The MathJax Consortium↵ *↕ -5.62 (#153) · ⚡ 0.1%
  14. ",") Act. Specifically, the suit accuses Yahoo↕ -5.59 (#3) · ⚡ 0.7%
  15. " closing" Alta Bates Hospital from closing, which is why↕ -5.59 (#7) · ⚡ 0.7%
  16. "else" test inside an #else clause.↵#else↕ -5.59 (#29) · ⚡ 0.3%
  17. "_only" able to modify read_only.↵the user↕ -5.53 (#4) · ⚡ 0.4%
  18. " instance"↵ * The singleton instance of this renderer class↕ -5.53 (#10) · ⚡ 0.5%
  19. " interim" will take over as interim chief, and the↕ -5.47 (#5) · ⚡ 0.7%
  20. " footage" 16mm footage of Cobain’s↕ -5.47 (#10) · ⚡ 0.5%
Qwen 3.5 9B Base · L17/C530

Country/state/region name

Confidence 2/5
Component profile — C530 · layer 17 · negative tail

Logit-lens tokens · negative

#1"及周边"#2"-wide"#3"乃至"#4"境内的"#5"-specific"#6"本土"#7"和中国"#8"独有的"#9"特有的"#10"に限"

R-lens tokens · negative

#1"本土"#2" residents"#3"及周边"#4"境内的"#5" citizens"#6" native"#7"本地的"#8" nationals"#9"-wide"#10"和中国"

Skewness

-3.07

Excess kurtosis

31.39rank #1451

Logcosh deviation

0.0854rank #531

Top-score occurrences · negative

  1. " Jersey" pension funds. New Jersey's top court overturned↕ -20.50 (#1) · ⚡ 10.6%
  2. " Zealand" change regime, New Zealand’s alpine habitats↕ -19.25 (#1) · ⚡ 9.1%
  3. " Louisiana" on a compromise between Louisiana legislators that allows state↕ -19.00 (#1) · ⚡ 9.7%
  4. " Zealand"015 New Zealand Prime Minister John Key↕ -18.88 (#1) · ⚡ 11.6%
  5. " Arkansas" concealed carry holders in Arkansas. One of the↕ -18.88 (#1) · ⚡ 8.2%
  6. " Arizona" First Things First for Arizona's Children.↵↵↕ -18.88 (#1) · ⚡ 7.6%
  7. " Delaware"For the latest on Delaware’s response, visit↕ -18.75 (#1) · ⚡ 8.5%
  8. " Wyoming" keynote speech, said Wyoming's a pretty attractive↕ -18.75 (#1) · ⚡ 9.4%
  9. " Kerala"↵↵Organisation : Information Kerala Mission (IKM↕ -18.25 (#1) · ⚡ 4.8%
  10. " Pakistan"21, became Pakistan's first transgender news↕ -18.12 (#1) · ⚡ 7.0%
  11. " Poland" will this mean for Poland’s LGBT+ people↕ -18.12 (#1) · ⚡ 7.0%
  12. " Texas" event was to promote Texas undergraduate research projects.↕ -18.00 (#1) · ⚡ 9.2%
  13. " Australia" a positive sign for Australia’s largest performing arts↕ -17.88 (#1) · ⚡ 8.1%
  14. " Pennsylvania"apturing customers from Pennsylvania who purchase wine and↕ -17.88 (#1) · ⚡ 8.9%
  15. " Utah", referencing a recent Utah Foundation report.↵↵↕ -17.75 (#1) · ⚡ 9.1%
  16. " Michigan" Richards is Coming to Michigan↵↵I’m excited↕ -17.62 (#1) · ⚡ 5.4%
  17. " Utah"'s largest ever in Utah — larger than↕ -17.62 (#1) · ⚡ 5.8%
  18. "orian"land.↵↵Victorian Council of Social Service↕ -17.62 (#1) · ⚡ 9.6%
  19. " Nigerian" characterization of headache among Nigerian healthcare workers or healthcare↕ -17.50 (#1) · ⚡ 6.6%
  20. " Leone"4 survey of Sierra Leonean radio listeners,↕ -17.38 (#1) · ⚡ 9.1%
Qwen 3.5 9B Base · L18/C3329

Mc- surname prefix

Confidence 5/5
Component profile — C3329 · layer 18 · positive tail

Logit-lens tokens · positive

#1"ane"#2"ք"#3"pu"#4"Leod"#5"王安石"#6"的气势"#7"amba"#8"大事"#9"ister"#10"انه"

R-lens tokens · positive

#1"爱尔兰"#2"Mac"#3" Ireland"#4" irland"#5"сполни"#6" Irish"#7" McA"#8"amba"#9"reland"#10"Austr"

Skewness

+1.32

Excess kurtosis

34.35rank #1695

Logcosh deviation

0.0215rank #3330

Top-score occurrences · positive

  1. " McN"↵↵So writes Mr McNair Wilson in his↕ +39.50 (#1) · ⚡ 36.7%
  2. "CG").↵↵TIM MCGRAW'S 1↕ +39.25 (#1) · ⚡ 38.7%
  3. " McL" and Jenny L. McLeroy of Eyster↕ +36.50 (#1) · ⚡ 33.2%
  4. " McN": Scotsman Allan McNish. Nish↕ +34.75 (#1) · ⚡ 28.1%
  5. " McC"↵Stephen P. McCarron, Louis↕ +34.25 (#1) · ⚡ 33.6%
  6. " McL" are comedians Josh McLane of Memphis,↕ +33.00 (#1) · ⚡ 28.5%
  7. " McDon" Eileen T. McDonough, Attorneys↕ +32.50 (#1) · ⚡ 28.3%
  8. "gu" directed to rmcguire@cpup↕ +32.25 (#1) · ⚡ 24.1%
  9. " McM"By Barbara E. McMullen & John F↕ +31.50 (#1) · ⚡ 27.1%
  10. " McM" cross country coach Joel McMullen the chance to↕ +31.00 (#1) · ⚡ 26.4%
  11. " Mc"No, this is McLaughlin, his↕ +30.25 (#1) · ⚡ 26.4%
  12. " McDon" in North Carolina; McDonnell Genome Institute at↕ +28.75 (#1) · ⚡ 26.2%
  13. " McC" HOLMES, and McCORD, Circuit Judges↕ +28.62 (#1) · ⚡ 32.0%
  14. "El"Charles R. McElwee, Robinson↕ +28.50 (#1) · ⚡ 28.9%
  15. " Mc" D. B. McReynolds[^↕ +27.88 (#1) · ⚡ 30.7%
  16. " Mc" his wife, Margaret McMurtrie. He↕ +27.62 (#1) · ⚡ 23.8%
  17. " McG"White House counsel Donald McGahn to have Mueller↕ +27.38 (#1) · ⚡ 27.7%
  18. " Mc" and Charles A. McCallum III of↕ +26.38 (#1) · ⚡ 26.6%
  19. "Gu" into Nigel (McGuinness). This is↕ +25.75 (#1) · ⚡ 20.2%
  20. " Mc"3, issued to McAlexander, White and↕ +25.62 (#1) · ⚡ 27.5%
Qwen 3.5 9B Base · L19/C146

Italian / Italy-related context

Confidence 4/5
Component profile — C146 · layer 19 · positive tail

Logit-lens tokens · positive

#1" significative"#2" splitted"#3" Moreover"#4"Moreover"#5" nowadays"#6" moreover"#7"enght"#8" indip"#9" lenght"#10"ò"

R-lens tokens · positive

#1" Italian"#2"意大利"#3" Italy"#4" italian"#5"イタリア"#6" Italians"#7"Italian"#8"Italy"#9" Italia"#10" италья"

Skewness

+8.09

Excess kurtosis

133.27rank #551

Logcosh deviation

0.1501rank #147

Top-score occurrences · positive

  1. " italian" ("More" in italian) that contains the↕ +34.75 (#1) · ⚡ 22.8%
  2. " can" ...↵↵As you can notice the name of↕ +34.25 (#1) · ⚡ 17.5%
  3. " is"ri.↵↵It is especially "the refusal↕ +32.00 (#1) · ⚡ 21.8%
  4. "ro" tab called "Altro" ("More"↕ +31.50 (#2) · ⚡ 12.6%
  5. " was" mentioned malonates was represented by a still↕ +31.00 (#1) · ⚡ 19.2%
  6. " focusing" in this field, focusing on the most important↕ +30.62 (#1) · ⚡ 20.3%
  7. " attempt" realistic. In the attempt to put some order↕ +30.50 (#1) · ⚡ 18.7%
  8. " was" diagnostic situation, it was decided to perform the↕ +30.12 (#1) · ⚡ 23.8%
  9. "6"-90-62-g00↕ +30.00 (#1) · ⚡ 19.6%
  10. " decided" organism. It was decided to schedule surgery to↕ +30.00 (#1) · ⚡ 18.8%
  11. "utto"79 anni Lutto nel mondo di musica↕ +29.50 (#1) · ⚡ 13.4%
  12. " Italy" of↵France and Italy. There are no↕ +29.25 (#1) · ⚡ 16.9%
  13. " also". This is confirmed also by the disclosure of↕ +29.00 (#1) · ⚡ 15.7%
  14. " due"A vital importance is due to the adhesive strength↕ +28.25 (#1) · ⚡ 15.3%
  15. " underline" treated. The Authors underline in this paper the↕ +28.25 (#1) · ⚡ 17.1%
  16. " problem" is ok. The problem is that i don↕ +28.12 (#1) · ⚡ 14.4%
  17. " light"2]).↵↵In light of these evidences↕ +27.88 (#1) · ⚡ 15.0%
  18. " quanto" le comunichiamo quanto segue:↕ +27.50 (#1) · ⚡ 16.4%
  19. " proved"ergic pathway has been proved to be fundamental in↕ +27.12 (#2) · ⚡ 13.6%
  20. " such" truncate the theory in such a way to also↕ +27.12 (#1) · ⚡ 16.8%
Qwen 3.5 9B Base · L19/C1472

HTML DOM access/manipulation

Confidence 4/5
Component profile — C1472 · layer 19 · negative tail

Logit-lens tokens · negative

#1" HTML"#2"HTML"#3" XHTML"#4" html"#5"元素"#6"_HTML"#7"Containing"#8" subtree"#9" markup"#10"TagName"

R-lens tokens · negative

#1" HTML"#2"HTML"#3" html"#4" XHTML"#5" Html"#6"html"#7"_html"#8" SVG"#9" CSS"#10" XML"

Skewness

-2.44

Excess kurtosis

23.88rank #2503

Logcosh deviation

0.0668rank #1473

Top-score occurrences · negative

  1. ">"'>More</button>↵ <div↕ -20.62 (#1) · ⚡ 11.5%
  2. ">">three</a>↵ </li↕ -20.00 (#1) · ⚡ 11.2%
  3. " :" 0 ? "" : e.childNodes[0↕ -19.12 (#1) · ⚡ 5.5%
  4. " same"multiple divs with same class on page)↕ -19.00 (#2) · ⚡ 7.8%
  5. "')."↵ $('.slide').each(function (i↕ -18.50 (#2) · ⚡ 8.0%
  6. " element" */ (element && element.firstChild &&↵ ↕ -18.38 (#1) · ⚡ 7.1%
  7. "\n" return null;↵ }↵↵ ↕ -17.88 (#1) · ⚡ 7.5%
  8. ")."El = $(this).parent();↵ ↕ -17.62 (#1) · ⚡ 6.2%
  9. " as" want to target all as that are inside #↕ -17.50 (#1) · ⚡ 5.5%
  10. ")."').html($(response).find("#created").↕ -17.50 (#1) · ⚡ 7.4%
  11. " document" var newContainer = document.createElement('div');↕ -17.25 (#1) · ⚡ 5.3%
  12. " root" be applied to the root element of components rendered↕ -17.12 (#1) · ⚡ 8.2%
  13. " is" the {{Element}} is not in the document↕ -17.00 (#1) · ⚡ 7.5%
  14. " element"↵↵ // Make element focusable↵ ↕ -17.00 (#2) · ⚡ 5.3%
  15. " =";↵↵var element = $("<img src="+↕ -16.88 (#1) · ⚡ 8.2%
  16. " within"-error class is not within form.checkout,↕ -16.62 (#2) · ⚡ 6.5%
  17. " current" {↵ return current;↵ }↕ -16.62 (#1) · ⚡ 6.2%
  18. " body"0];↵ body.innerHTML = '<div↕ -16.62 (#3) · ⚡ 6.7%
  19. " next" - will return the next element of the parent↕ -16.50 (#1) · ⚡ 5.4%
  20. ".dom"item = $K.dom('<a>');↵↕ -16.50 (#2) · ⚡ 4.1%
Qwen 3.5 9B Base · L19/C2291

?

Confidence 0/5
Component profile — C2291 · layer 19 · positive tail

Logit-lens tokens · positive

#1"ین"#2".gov"#3"äßig"#4"://"#5".exe"#6".mov"#7"和我的"#8"opodob"#9"大观"#10"\\system"

R-lens tokens · positive

#1" isnt"#2" doesnt"#3" america"#4" microsoft"#5" thats"#6" didnt"#7" torino"#8" english"#9" rotterdam"#10"->"

Skewness

+1.85

Excess kurtosis

17.55rank #2851

Logcosh deviation

0.0508rank #2292

Top-score occurrences · positive

  1. " parm"↵↵80g parmigiano regg↕ +22.88 (#1) · ⚡ 10.5%
  2. "java"2： 提升java性能的基本方法↵↕ +22.50 (#1) · ⚡ 8.0%
  3. " asp" service?↵↵My asp.net site coded in↕ +22.00 (#1) · ⚡ 10.5%
  4. " easy".↵↵Sew easy, which begins its↕ +20.00 (#1) · ⚡ 8.0%
  5. " react" I do it in react native?↵↵A↕ +18.12 (#1) · ⚡ 8.8%
  6. "apache" ,big data ,apache spark ,cfs↕ +18.12 (#1) · ⚡ 6.9%
  7. " kafka".debug("Creating new kafka producer using brokers:↕ +17.75 (#1) · ⚡ 8.2%
  8. " genes" for salesforce with genesys. ↵Let↕ +17.62 (#1) · ⚡ 5.9%
  9. " im"! meeting them through imac... that is↕ +17.50 (#1) · ⚡ 7.9%
  10. " tan" than fun. But tan Tui is so↕ +17.50 (#1) · ⚡ 7.9%
  11. " yas" singed mains, yasuo mains, ann↕ +17.38 (#1) · ⚡ 6.4%
  12. " high" background color of this highcharts chart:↵↕ +17.25 (#1) · ⚡ 7.3%
  13. " z"inding meal prices in zimbabwe↵↵Grinding↕ +17.25 (#1) · ⚡ 7.0%
  14. "zion" were more christian "zionists" than jewish↕ +17.25 (#2) · ⚡ 5.3%
  15. " halo" after this game of halo."↕ +17.25 (#1) · ⚡ 5.2%
  16. " x" create fictional designs in x-plane for which no↕ +17.12 (#1) · ⚡ 9.0%
  17. "pc"Encode messages from ldpc parity check matrix in↕ +17.12 (#2) · ⚡ 3.8%
  18. " fund" they received from the fund. They turned to↕ +17.00 (#1) · ⚡ 7.3%
  19. "anna"misc. photos↵↵anna's photos↵↵BA↕ +17.00 (#1) · ⚡ 5.7%
  20. " i" this chip over the i2c bus.↕ +17.00 (#1) · ⚡ 9.3%
Qwen 3.5 9B Base · L19/C3174

?

Confidence 0/5
Component profile — C3174 · layer 19 · positive tail

Logit-lens tokens · positive

#1" genuinely"#2" fundamentally"#3"之所以"#4"权衡"#5"我觉得"#6" принципи"#7"但其实"#8"锐意"#9"这件事情"#10" FUCK"

R-lens tokens · positive

#1" fundamentally"#2" genuinely"#3" FUCK"#4" принципи"#5" shitty"#6" objectively"#7"fuck"#8"我觉得"#9"Fuck"#10" fucking"

Skewness

+0.08

Excess kurtosis

4.05rank #3580

Logcosh deviation

0.0337rank #3175

Top-score occurrences · positive

  1. " if" I wonder every day if your crazy family is↕ +12.88 (#3) · ⚡ 2.7%
  2. " But" 30! But that’s okay too↕ +12.00 (#6) · ⚡ 2.5%
  3. " I" grown-up stuff and I'm stressed and I↕ +11.94 (#2) · ⚡ 3.4%
  4. "." believing even in Purpose. And that scared the↕ +11.25 (#2) · ⚡ 2.5%
  5. " you". I'm scared you're going to change↕ +10.88 (#2) · ⚡ 2.4%
  6. " And" even in Purpose. And that scared the shit↕ +10.81 (#2) · ⚡ 2.2%
  7. " realised" moment today when I realised that actually no,↕ +10.56 (#4) · ⚡ 2.0%
  8. "And" hurt Henrik.”↵↵And I don’t mean↕ +10.44 (#3) · ⚡ 2.0%
  9. "y" bad or whiney or entitled for asking↕ +10.19 (#5) · ⚡ 1.5%
  10. "?" know the worst part? For a good ↕ +10.00 (#5) · ⚡ 1.3%
  11. " that" in Purpose. And that scared the shit out↕ +9.88 (#3) · ⚡ 2.0%
  12. "\n\n". and sad.↵↵sad because essentially what↕ +9.88 (#2) · ⚡ 2.1%
  13. " Now" on that idea. Now I can truly say↕ +9.75 (#2) · ⚡ 1.6%
  14. "." have to believe anything. It means that I↕ +9.69 (#3) · ⚡ 2.0%
  15. "." be a better person. We can all figure↕ +9.62 (#5) · ⚡ 1.6%
  16. "," that back to myself, however, I realize↕ +9.62 (#2) · ⚡ 1.9%
  17. "," to myself, however, I realize just how↕ +9.56 (#1) · ⚡ 2.4%
  18. " was" know what? It was liberating.↵↵↕ +9.31 (#4) · ⚡ 2.2%
  19. " about" did so. Think about how absurd that is↕ +9.31 (#5) · ⚡ 1.1%
  20. " is" realize how absurd it is that if another Chris↕ +9.25 (#4) · ⚡ 1.4%
Qwen 3.5 9B Base · L20/C352

Both

Confidence 5/5
Component profile — C352 · layer 20 · negative tail

Logit-lens tokens · negative

#1" sides"#2" sexes"#3"/all"#4" halves"#5" ends"#6" lados"#7"uding"#8"rops"#9"çinin"#10" genders"

R-lens tokens · negative

#1" sides"#2" sexes"#3"这两种"#4" halves"#5"/all"#6"这两个"#7" extremes"#8" genders"#9" lados"#10" ends"

Skewness

-10.24

Excess kurtosis

283.23rank #229

Logcosh deviation

0.1205rank #353

Top-score occurrences · negative

  1. " Both" 3 weeks. Both the rats’ behavior↕ -42.50 (#1) · ⚡ 30.5%
  2. " Both" creative nonfiction. Both graduate and or flagship↕ -42.25 (#1) · ⚡ 37.5%
  3. " both" copies, and that both that copyright notice↵↕ -42.25 (#1) · ⚡ 19.2%
  4. "Both"nostic software.↵↵Both projects are great,↕ -38.75 (#1) · ⚡ 32.6%
  5. " Both" to the attack. Both leave the dinner uncon↕ -38.50 (#1) · ⚡ 31.6%
  6. " Both" reliable prime minister. Both choices are of seminal↕ -38.25 (#1) · ⚡ 32.2%
  7. " both" use it to make both advertising and content more↕ -38.00 (#1) · ⚡ 15.6%
  8. " both" it turns out, both of these issues were↕ -37.50 (#1) · ⚡ 23.9%
  9. " Both"entulous patients. Both cases presented here demonstrated↕ -37.50 (#1) · ⚡ 37.7%
  10. " both" irrelevant.↵↵Yet both in Syria and Iraq↕ -37.25 (#1) · ⚡ 36.7%
  11. "Both" this manner.↵Both Moore, and Abdel↕ -37.25 (#1) · ⚡ 35.4%
  12. " Both" device performance significantly. Both the effective increase in↕ -37.00 (#1) · ⚡ 34.2%
  13. " both", will contribute to both of these aspirations,↕ -36.75 (#1) · ⚡ 30.9%
  14. " both" was used to measure both the mass and charge↕ -36.50 (#1) · ⚡ 32.4%
  15. " both" reporting positive relationships for both groups \[[@↕ -36.50 (#1) · ⚡ 28.9%
  16. " Both") firestorms. Both wind and plume↕ -36.50 (#1) · ⚡ 33.6%
  17. " Both"rycians. Both Mariandynians↕ -36.25 (#1) · ⚡ 27.1%
  18. " Both" an open fire. Both of our plates were↕ -36.25 (#1) · ⚡ 26.4%
  19. "Both"ato cakes.”↵↵Both Hasseena and↕ -35.75 (#1) · ⚡ 35.5%
  20. "Both"'s hands.↵↵Both the Conservatives and Liberals↕ -35.75 (#1) · ⚡ 33.2%
Qwen 3.5 9B Base · L20/C1145

Related to letter F or expecting F

Confidence 1/5
Component profile — C1145 · layer 20 · positive tail

Logit-lens tokens · positive

#1" f"#2"\tf"#3"_f"#4",f"#5".f"#6" ф"#7"$f"#8" الف"#9" फ"#10"\"f"

R-lens tokens · positive

#1" f"#2"\tf"#3",f"#4"$f"#5"_f"#6".f"#7" F"#8" ফ"#9" फ"#10" الف"

Skewness

+2.76

Excess kurtosis

23.25rank #2820

Logcosh deviation

0.0783rank #1146

Top-score occurrences · positive

  1. "\n" Weiss (CH)↵ Flavio Favelli↕ +22.38 (#1) · ⚡ 8.3%
  2. "angi" Park↵↵Aorangi Forest Park is a↕ +19.88 (#1) · ⚡ 8.4%
  3. "Stone"<|endoftext|>Stone flaming↵↵Stone flaming or thermaling↕ +19.75 (#1) · ⚡ 7.4%
  4. "string" <key> <string>form_id</↕ +19.00 (#2) · ⚡ 6.1%
  5. "ro" nitrofirenitrofire Watch↵↵5↕ +18.88 (#1) · ⚡ 8.5%
  6. "CR" (FAO [@CR20]). It↕ +17.88 (#3) · ⚡ 5.5%
  7. ")" (AF/AU)↵ Federico Bacci↕ +17.62 (#2) · ⚡ 5.7%
  8. " STAT"46↵#define STATFS_FUTEX↕ +17.62 (#1) · ⚡ 8.4%
  9. "2" KNN 42.f recovered from an↕ +17.62 (#1) · ⚡ 6.4%
  10. "loader"otfload/fontloader-fonts-ext.lua↕ +17.50 (#2) · ⚡ 5.6%
  11. " structures"Fires within building structures often start in a↕ +17.38 (#1) · ⚡ 7.3%
  12. "='" <div class='fl pad1y↕ +17.25 (#2) · ⚡ 3.9%
  13. " $"↵ "$finish $fmonitorb $↕ +17.25 (#1) · ⚡ 6.9%
  14. " class"↵ <div class='fl pad1↕ +17.00 (#3) · ⚡ 4.6%
  15. ")" & Weiss (CH)↵ Flavio Fav↕ +17.00 (#1) · ⚡ 4.6%
  16. "clarations"no-deprecated-declarations -fstack-↕ +16.88 (#1) · ⚡ 8.3%
  17. "o" Finland↵↵Flamingo Entertainment Centre is the↕ +16.88 (#2) · ⚡ 4.4%
  18. "-"synth_lena-ffvhuff4↕ +16.62 (#1) · ⚡ 5.9%
  19. ")" & Alexander (IT)↵ Fahim Amir↕ +16.50 (#2) · ⚡ 4.2%
  20. " of" :D↵World of Final Fantasy is a↕ +16.38 (#1) · ⚡ 5.8%
Qwen 3.5 9B Base · L20/C2259

Document/source attribution

Confidence 4/5
Component profile — C2259 · layer 20 · negative tail

Logit-lens tokens · negative

#1"的一份"#2" accompanying"#3" documents"#4" filings"#5" excerpts"#6"发送给"#7" emailed"#8" submitted"#9" penned"#10" recent"

R-lens tokens · negative

#1" documents"#2" posted"#3" filings"#4" emailed"#5" excerpts"#6" document"#7" penned"#8" written"#9" published"#10"的一份"

Skewness

-2.72

Excess kurtosis

42.52rank #1977

Logcosh deviation

0.0557rank #2260

Top-score occurrences · negative

  1. " audio"↵↵ADVERTISEMENT↵↵In audio of the 9↕ -28.50 (#1) · ⚡ 15.9%
  2. " in" the Net movement say in their video (above↕ -27.50 (#1) · ⚡ 22.3%
  3. " sheet" according to the term sheet detailing the loan.↕ -27.12 (#1) · ⚡ 14.1%
  4. " letter"aptcha↵↵In a letter to MSU,↕ -26.75 (#2) · ⚡ 13.6%
  5. " email".↵↵In an email, the organisers said↕ -26.62 (#1) · ⚡ 15.3%
  6. " a" endorsement.↵↵In a letter to the Department↕ -26.50 (#1) · ⚡ 19.5%
  7. " a"inkley says in a statement the lien was↕ -25.88 (#1) · ⚡ 19.0%
  8. " in"Wire,” says Patrick in the press release.↕ -25.75 (#1) · ⚡ 20.5%
  9. " a", prosecutors noted in a motion that has been↕ -25.38 (#1) · ⚡ 20.0%
  10. " in"DO Mobile, said in a statement about the↕ -25.38 (#1) · ⚡ 20.5%
  11. " in" Transportation Committee, stated in his correspondence to Samp↕ -25.00 (#1) · ⚡ 20.2%
  12. " motion" his attorneys, a motion unsealed Wednesday with↕ -25.00 (#1) · ⚡ 13.8%
  13. " memorandum" also wrote in a memorandum Thursday, April ↕ -24.88 (#1) · ⚡ 13.3%
  14. " application".↵↵In the application, Coastal GasLink↕ -24.75 (#1) · ⚡ 12.4%
  15. " a" O'Connor said in a statement:↵↵"We↕ -24.62 (#1) · ⚡ 17.3%
  16. " in" Thomas O'Connor said in a statement:↵↵↕ -24.62 (#1) · ⚡ 17.7%
  17. " in" ban Trump, said in response to the petition↕ -24.50 (#1) · ⚡ 17.6%
  18. " in",” Kroger wrote in an email. “↕ -24.38 (#1) · ⚡ 19.1%
  19. " an"ampel wrote in an email to Nassar↕ -24.00 (#1) · ⚡ 18.2%
  20. " release"anza in a press release. “We’ll↕ -23.75 (#1) · ⚡ 13.9%
Qwen 3.5 9B Base · L21/C704

Microsoft platform technical documentation

Confidence 4/5
Component profile — C704 · layer 21 · negative tail

Logit-lens tokens · negative

#1" Microsoft"#2"微软"#3"Microsoft"#4".microsoft"#5"icrosoft"#6" microsoft"#7" Windows"#8"Windows"#9"microsoft"#10" PowerShell"

R-lens tokens · negative

#1" Microsoft"#2"微软"#3"Microsoft"#4".microsoft"#5"icrosoft"#6" microsoft"#7" Windows"#8"Windows"#9"microsoft"#10"微軟"

Skewness

-3.94

Excess kurtosis

46.69rank #2047

Logcosh deviation

0.0964rank #705

Top-score occurrences · negative

  1. "\n\n"/permissions-reference).↵↵|Permission type|↕ -29.50 (#1) · ⚡ 8.3%
  2. " API" required to call this API. To learn more↕ -26.62 (#1) · ⚡ 6.7%
  3. "---" apiPageType↵---↵↵# Delete android↕ -25.62 (#1) · ⚡ 5.5%
  4. "\n"↵↵## Request headers↵|Header|Value↕ -25.38 (#2) · ⚡ 7.2%
  5. " headers"```↵↵## Request headers↵|Header|↕ -24.75 (#2) · ⚡ 6.3%
  6. "." to call this API. To learn more,↕ -24.62 (#1) · ⚡ 6.4%
  7. "|"## Request headers↵|Header|Value|↕ -24.50 (#2) · ⚡ 6.2%
  8. "." example of the request.↵``` http↵↕ -24.38 (#1) · ⚡ 6.1%
  9. "\n\n".ReadWrite.All|↵↵## HTTP Request↵↕ -24.38 (#1) · ⚡ 7.4%
  10. "\n"Header|Value|↵|:---|↕ -23.50 (#3) · ⚡ 5.2%
  11. ".graph"↵↵Namespace: microsoft.graph↵↵> **Important↕ -23.50 (#1) · ⚡ 7.7%
  12. "|"---|:---|↵|Authorization|↕ -23.50 (#2) · ⚡ 6.2%
  13. "-->""↵}↵-->↵``` http↵↕ -23.25 (#1) · ⚡ 6.5%
  14. ":**"↵↵> **Important:** Microsoft Graph APIs under↕ -23.25 (#1) · ⚡ 6.2%
  15. " Request"## Example↵↵### Request↵Here is an↕ -22.88 (#1) · ⚡ 6.7%
  16. "\n\n"|application/json|↵↵## Request body↵↕ -22.38 (#1) · ⚡ 7.6%
  17. ":"StoreApp↵↵Namespace: microsoft.graph↵↵>↕ -22.38 (#1) · ⚡ 8.2%
  18. "\n" "intune"↵doc_type: api↕ -22.25 (#1) · ⚡ 5.2%
  19. " more" API. To learn more, including how to↕ -22.00 (#1) · ⚡ 4.8%
  20. "beta" APIs under the /beta version are subject to↕ -22.00 (#2) · ⚡ 5.9%
Qwen 3.5 9B Base · L22/C2808

Urination / bladder function

Confidence 4/5
Component profile — C2808 · layer 22 · negative tail

Logit-lens tokens · negative

#1"排尿"#2" bladder"#3" toilet"#4"尿道"#5" urinary"#6" Toilet"#7"膀胱"#8" urine"#9"泌尿"#10" urin"

R-lens tokens · negative

#1" toilet"#2"排尿"#3" urinary"#4" urine"#5" bladder"#6" Toilet"#7" toilets"#8" urin"#9"膀胱"#10"尿道"

Skewness

-2.18

Excess kurtosis

28.84rank #2956

Logcosh deviation

0.0521rank #2809

Top-score occurrences · negative

  1. " and"continence are frequent and urgent mictur↕ -31.50 (#1) · ⚡ 15.2%
  2. " with" micturitions with incontinence.↕ -30.88 (#1) · ⚡ 15.1%
  3. " and", decreased sensation, and poor emptying ([↕ -26.50 (#1) · ⚡ 16.4%
  4. "itions" urgent micturitions with incontinence↕ -26.50 (#1) · ⚡ 12.2%
  5. " poor" decreased sensation, and poor emptying ([@↕ -26.38 (#1) · ⚡ 14.6%
  6. " urine" are urine storage and urine disposal, and the↕ -24.38 (#1) · ⚡ 12.9%
  7. " involuntary" condition in which an involuntary loss of urine occurs↕ -24.00 (#1) · ⚡ 9.0%
  8. "ec" the desire to defecate. This urge↕ -23.00 (#1) · ⚡ 11.1%
  9. " desire"al walls stimulate the desire to defecate↕ -22.88 (#1) · ⚡ 8.1%
  10. " and" bladder are urine storage and urine disposal, and↕ -22.88 (#1) · ⚡ 13.5%
  11. "ec" involuntary control↵Defecation may be involuntary↕ -22.62 (#1) · ⚡ 9.1%
  12. " by", may be caused by physical injury, nerve↕ -22.50 (#1) · ⚡ 10.2%
  13. "uria", such as nocturia, urgency, and↕ -22.50 (#1) · ⚡ 10.7%
  14. " bladder"17% impaired bladder compliance. Half had↕ -22.38 (#2) · ⚡ 10.4%
  15. " of". Ten patients complained of residual lower urinary tract↕ -22.25 (#1) · ⚡ 12.8%
  16. " by"ation is normally assisted by taking a deep breath↕ -22.25 (#1) · ⚡ 9.1%
  17. " ur" and the need to urinate often or urgently↕ -22.12 (#1) · ⚡ 8.3%
  18. "." anal sphincter. If the urge is↕ -22.12 (#1) · ⚡ 10.3%
  19. " urinary"ing symptoms, and urinary flow were minor.↕ -21.88 (#1) · ⚡ 10.7%
  20. " of" as the sole cause of their incontinence↕ -21.38 (#1) · ⚡ 13.0%
Qwen 3.5 9B Base · L22/C3385

Right-wing / conservative political context

Confidence 4/5
Component profile — C3385 · layer 22 · negative tail

Logit-lens tokens · negative

#1"保守"#2" conservative"#3" conservatives"#4"dry"#5" Conservative"#6"点击右上角"#7" dry"#8" 보수"#9"右上"#10"右上角"

R-lens tokens · negative

#1"保守"#2" conservative"#3" Conservative"#4" 보수"#5" conservatism"#6" conservatives"#7"dry"#8" консер"#9"点击右上角"#10"右上"

Skewness

-0.21

Excess kurtosis

7.65rank #3813

Logcosh deviation

0.0390rank #3386

Top-score occurrences · negative

  1. " for" battle against Obama and for America must continue,↕ -15.62 (#1) · ⚡ 6.5%
  2. " the" this area rather than the libertarians from the↕ -15.00 (#1) · ⚡ 6.8%
  3. " a" had pledged to nominate a Southerner as↕ -14.56 (#1) · ⚡ 5.9%
  4. " of" over the steady objection of Republican lawmakers.↵↵↕ -14.44 (#1) · ⚡ 7.8%
  5. " armed" the West supported and armed right-wing forces that↕ -14.44 (#1) · ⚡ 4.6%
  6. " Of"ized Social Security? Of the merits of ↕ -14.00 (#2) · ⚡ 4.5%
  7. " favor" Right-libertarians favor private property—i↕ -14.00 (#3) · ⚡ 3.2%
  8. " and" responsible for privatizations and heavy spending cuts in↕ -14.00 (#2) · ⚡ 5.6%
  9. " against" Northam for voting against a bill that would↕ -13.56 (#3) · ⚡ 4.8%
  10. " to" which has been linked to the Tea Party --↕ -13.50 (#1) · ⚡ 6.3%
  11. " a"arming Policy Foundation, a climate-change denying organization↕ -13.38 (#1) · ⚡ 4.8%
  12. " for" KKK and fight for race and nation,”↕ -13.19 (#1) · ⚡ 4.3%
  13. " on"Thank You" tour on states he won,↕ -13.00 (#1) · ⚡ 4.4%
  14. " in" Reagan-Thatcher faith in unregulated capitalism,↕ -12.75 (#2) · ⚡ 2.6%
  15. " censor" of conspiring to censor conservative content, notably↕ -12.44 (#4) · ⚡ 3.4%
  16. "," and their guns alone, homeschoolers who wish↕ -12.31 (#3) · ⚡ 4.0%
  17. " the"-class Brexit voters and the privately educated opportunists↕ -12.31 (#1) · ⚡ 5.1%
  18. "-fat"-fat vegan to high-fat, low-carb↕ -12.19 (#2) · ⚡ 3.6%
  19. "4" Democrats and 64 by Republicans. Republicans↕ -12.06 (#2) · ⚡ 3.4%
  20. " Scalia" days by Justices Scalia and Thomas), who↕ -12.06 (#4) · ⚡ 3.8%
Qwen 3.5 9B Base · L23/C1583

Limit

Confidence 5/5
Component profile — C1583 · layer 23 · negative tail

Logit-lens tokens · negative

#1" imposed"#2"rophe"#3"民事行为能力"#4"定了"#5"定的"#6" sayıda"#7"Reached"#8"lessly"#9"条件下的"#10"anei"

R-lens tokens · negative

#1" imposed"#2"民事行为能力"#3"rophe"#4" enforced"#5" restrict"#6"定了"#7" restricting"#8"Reached"#9"(limit"#10" restricted"

Skewness

-4.81

Excess kurtosis

111.47rank #907

Logcosh deviation

0.0707rank #1584

Top-score occurrences · negative

  1. " limit"ular specification of the limit for the power over↕ -39.75 (#1) · ⚡ 26.0%
  2. " Limit" Yoshizaki's Limit Gunz PVC Statue↕ -38.00 (#1) · ⚡ 26.6%
  3. " limit". There is a limit on how fast tun↕ -37.50 (#1) · ⚡ 17.8%
  4. " limit" means there is a limit.↵↵Since there↕ -36.00 (#1) · ⚡ 20.3%
  5. " limits" up ideas of monthly limits or added expenses.↕ -35.25 (#1) · ⚡ 22.0%
  6. " Limit" a simple principle. Limit the area very accurately↕ -35.00 (#1) · ⚡ 24.5%
  7. " Limit" have been contemplated. Limitations could be amended↕ -35.00 (#1) · ⚡ 24.3%
  8. " limit" it's not a limit! Each additional GB↕ -34.25 (#1) · ⚡ 18.8%
  9. " limited" for Ron Paul and limited government conservatives at a↕ -34.25 (#1) · ⚡ 22.0%
  10. " limit" the world.The mailing limit for Christmas presents is↕ -33.50 (#1) · ⚡ 18.5%
  11. " Limit", or Alternatively, Limitation of Liability.↕ -33.25 (#1) · ⚡ 21.2%
  12. " limited" spreads, but is limited by the fact that↕ -33.00 (#1) · ⚡ 17.1%
  13. " limit"Since there is a limit, it is best↕ -32.75 (#1) · ⚡ 18.1%
  14. "Limit" – VIP Ticket (Limit Quantities)↵↵↕ -32.75 (#1) · ⚡ 23.4%
  15. " limited"↵This method is limited by the gable↕ -32.50 (#1) · ⚡ 21.7%
  16. " limited"ahat Dapat’ limited edition red scarf that↕ -32.50 (#1) · ⚡ 17.9%
  17. " limited" in your inbox is limited to only contacts you↕ -32.25 (#1) · ⚡ 16.9%
  18. " limitations".↵There are limitations on the amount of↕ -32.00 (#1) · ⚡ 19.5%
  19. " limit"! There is a limit of 25↕ -31.38 (#1) · ⚡ 15.9%
  20. " limited" sleeve. The special limited-edition box was↕ -30.75 (#1) · ⚡ 15.8%
Qwen 3.5 9B Base · L25/C602

Lawsuit / civil litigation

Confidence 5/5
Component profile — C602 · layer 25 · negative tail

Logit-lens tokens · negative

#1" lawsuit"#2" lawsuits"#3" litigation"#4" suits"#5"起诉"#6" suit"#7" suing"#8" plaintiffs"#9" claims"#10"诉讼"

R-lens tokens · negative

#1" lawsuit"#2" lawsuits"#3" litigation"#4" claims"#5"起诉"#6" suits"#7"诉讼"#8" plaintiffs"#9" suing"#10" sued"

Skewness

-5.22

Excess kurtosis

96.00rank #1148

Logcosh deviation

0.1019rank #603

Top-score occurrences · negative

  1. " this" Central also filed this action for damages in↕ -35.75 (#1) · ⚡ 28.5%
  2. " this" longer a party to this case, Dr.↕ -33.25 (#1) · ⚡ 26.0%
  3. " instant" and plaintiff instituted the instant action seeking a refund↕ -32.75 (#1) · ⚡ 23.4%
  4. "This".↵↵Background↵This Court entered judgment in↕ -32.50 (#1) · ⚡ 24.1%
  5. " the" the refiling of the suit if the district↕ -32.50 (#1) · ⚡ 26.2%
  6. " the" as of the time the complaint is filed.↕ -32.25 (#1) · ⚡ 21.8%
  7. " his"from the dismissal of his civil rights action.↕ -32.00 (#1) · ⚡ 22.3%
  8. " the", to abate the action against it;↕ -31.50 (#1) · ⚡ 26.4%
  9. " this"), Kennedy argues that this court has no subject↕ -31.00 (#1) · ⚡ 17.5%
  10. " this" accident giving rise to this action occurred when plaintiff↕ -30.50 (#1) · ⚡ 20.7%
  11. " the" limitations would otherwise bar the refiling of the↕ -30.50 (#1) · ⚡ 23.9%
  12. " was" upon federal diversity jurisdiction was filed in this Court↕ -30.25 (#1) · ⚡ 17.0%
  13. " this", the City began this action and sought a↕ -30.12 (#1) · ⚡ 19.4%
  14. "arty" defendant in a multiparty lawsuit extinguishes the↕ -30.00 (#1) · ⚡ 15.9%
  15. "'" order dismissing appellants' second amended complaint with↕ -29.62 (#1) · ⚡ 21.4%
  16. " this"ride Offshore filed this complaint on October ↕ -29.50 (#1) · ⚡ 22.1%
  17. " suit". BACKGROUND↵↵This suit arises from the attempted↕ -29.50 (#1) · ⚡ 18.5%
  18. " action" second, whether this action is barred by the↕ -29.50 (#1) · ⚡ 13.9%
  19. " the"). Plaintiff then filed the instant action in federal↕ -29.38 (#1) · ⚡ 18.5%
  20. " his" prohibiting Pace from pursuing his state claim and the↕ -29.38 (#1) · ⚡ 20.1%
Qwen 3.5 9B Base · L25/C1260

Maryland / Baltimore

Confidence 4/5
Component profile — C1260 · layer 25 · positive tail

Logit-lens tokens · positive

#1" Maryland"#2" Baltimore"#3" MD"#4" Md"#5"Md"#6"MD"#7"md"#8" md"#9"altimore"#10".md"

R-lens tokens · positive

#1" Maryland"#2" Baltimore"#3" Md"#4" MD"#5"altimore"#6"Md"#7"MD"#8"md"#9" md"#10"/md"

Skewness

+7.15

Excess kurtosis

231.30rank #309

Logcosh deviation

0.0792rank #1261

Top-score occurrences · positive

  1. " in"the farm"] located in Talbot County,↕ +64.00 (#1) · ⚡ 42.2%
  2. ",", Annapolis, and Washington, DC↕ +50.50 (#1) · ⚡ 37.3%
  3. " from" less than five miles from the Baltimore - Washington↕ +49.75 (#1) · ⚡ 35.4%
  4. "," destinations such as Baltimore, Annapolis,↕ +48.50 (#1) · ⚡ 41.6%
  5. " Maryland" Court, D. Maryland.↵October ↕ +45.75 (#1) · ⚡ 41.4%
  6. " such" many major tourist destinations such as Baltimore, An↕ +45.50 (#1) · ⚡ 32.6%
  7. ","↵Riverdale Park, Maryland, a town↕ +44.75 (#1) · ⚡ 36.7%
  8. " in" Maryland School of Medicine in Baltimore and colleagues studied↕ +44.25 (#1) · ⚡ 32.2%
  9. " of"9.↵Court of Special Appeals of Maryland↕ +43.50 (#1) · ⚡ 29.3%
  10. " and" Routes 32 and 198↕ +43.00 (#1) · ⚡ 29.1%
  11. " attorney"↵↵Yesterday state's attorney Scott D. Sh↕ +41.75 (#1) · ⚡ 37.3%
  12. " off" which can be accessed off of West Nursery Road↕ +41.75 (#1) · ⚡ 28.3%
  13. "."1967.↵Court of Special↕ +41.00 (#1) · ⚡ 35.2%
  14. " Code". See Md. Code Ann., Cts↕ +41.00 (#1) · ⚡ 28.3%
  15. " " Turover, 195 Md↕ +41.00 (#1) · ⚡ 35.7%
  16. " at" been a faculty member at the University of Maryland↕ +40.25 (#1) · ⚡ 23.3%
  17. " to" lot of fun driving to Baltimore. The George↕ +39.75 (#1) · ⚡ 28.7%
  18. " ".↵No. 102,↕ +39.50 (#1) · ⚡ 29.5%
  19. " with"Blake, in conjunction with↵ Police Commissioner↕ +39.25 (#1) · ⚡ 29.5%
  20. " alternate" driver to use our alternate entrance, which can↕ +39.00 (#1) · ⚡ 22.3%
Qwen 3.5 9B Base · L25/C2695

Tell

Confidence 5/5
Component profile — C2695 · layer 25 · positive tail

Logit-lens tokens · positive

#1" story"#2" stories"#3"故事"#4"的故事"#5"故事的"#6" Story"#7"Story"#8"story"#9" tales"#10" Stories"

R-lens tokens · positive

#1" story"#2" stories"#3"故事"#4"的故事"#5" Story"#6"故事的"#7"Story"#8"story"#9"这个故事"#10" Stories"

Skewness

+3.42

Excess kurtosis

76.78rank #1458

Logcosh deviation

0.0520rank #2696

Top-score occurrences · positive

  1. " telling" progression, they are telling you that it’s↕ +35.25 (#1) · ⚡ 25.4%
  2. " tell" sheet and the stats tell you about your character↕ +34.75 (#1) · ⚡ 21.7%
  3. " tell" domination. I can tell you that Cathery↕ +33.50 (#1) · ⚡ 19.4%
  4. " tell" don’t want to tell the reader what they↕ +33.50 (#1) · ⚡ 24.1%
  5. " tells" so far this summer tells me people want some↕ +33.25 (#1) · ⚡ 24.3%
  6. " tell" numbers alone don’t tell the tale of Hay↕ +33.00 (#1) · ⚡ 18.1%
  7. " tells" asks why and she tells him that she didn↕ +33.00 (#1) · ⚡ 23.6%
  8. " tell"early-warning creatures that tell us when things are↕ +33.00 (#1) · ⚡ 19.7%
  9. " tell" Now, let me tell you something: Sara↕ +33.00 (#1) · ⚡ 21.8%
  10. " tells" customers. Ilja tells of finding phone numbers↕ +32.75 (#1) · ⚡ 21.8%
  11. " tell", no challenge will tell us everything about how↕ +32.50 (#1) · ⚡ 17.7%
  12. " telling" an unnamed male narrator telling the story of how↕ +31.75 (#1) · ⚡ 24.2%
  13. " telling" were gonna sit around telling funny stories about our↕ +31.38 (#1) · ⚡ 16.3%
  14. " tell" people the space to tell their stories across the↕ +31.12 (#1) · ⚡ 22.9%
  15. " tells"man. Arthen tells me that in this↕ +31.00 (#1) · ⚡ 21.4%
  16. " tell" milkshake expert will tell you that a milk↕ +31.00 (#1) · ⚡ 17.2%
  17. " tell" building – some that tell of Shah Jahan↕ +31.00 (#1) · ⚡ 22.1%
  18. " told" of issues, Mitchell told the story of a↕ +31.00 (#1) · ⚡ 23.9%
  19. " tell"acihome needs to tell people he is a↕ +30.88 (#1) · ⚡ 23.3%
  20. " tells" popular. Agency theory tells us that a principal↕ +30.62 (#1) · ⚡ 23.2%
Qwen 3.5 9B Base · L25/C3939

Causal graphical-model reasoning

Confidence 3/5
Component profile — C3939 · layer 25 · negative tail

Logit-lens tokens · negative

#1"��"#2"/lic"#3"льд"#4"众所周知"#5"mej"#6"oric"#7" caus"#8" duyên"#9"iep"#10"媳"

R-lens tokens · negative

#1"��"#2" causal"#3"因果关系"#4"/lic"#5" caus"#6"因果"#7"ugin"#8"tabpanel"#9" Dependencies"#10"OSP"

Skewness

-0.70

Excess kurtosis

18.01rank #3496

Logcosh deviation

0.0167rank #3940

Top-score occurrences · negative

  1. " @"6graphical; @Spirtes2↕ -37.50 (#1) · ⚡ 13.7%
  2. "[@" have been discussed [@Zhang20↕ -37.00 (#1) · ⚡ 14.8%
  3. " the" DAG framework is the *causal Mark↕ -36.25 (#1) · ⚡ 13.2%
  4. "[@" polynomial-time algorithms [@Glymour1↕ -35.00 (#1) · ⚡ 13.0%
  5. " *" framework is the *causal Markov↕ -32.75 (#1) · ⚡ 12.2%
  6. ";"1987; @Spirtes↕ -31.62 (#1) · ⚡ 13.7%
  7. " P"ality assumptions, the P-minimality and↕ -28.88 (#1) · ⚡ 13.3%
  8. " of" we address the issue of identifiability for↕ -27.75 (#2) · ⚡ 10.7%
  9. "st" the recently introduced sparsest Markov representation (↕ -27.25 (#1) · ⚡ 10.6%
  10. " sparse", the recently introduced sparsest Markov representation↕ -26.62 (#1) · ⚡ 9.3%
  11. ";"1987; @Spirtes↕ -26.50 (#1) · ⚡ 10.8%
  12. "G"-time algorithms [@Glymour19↕ -26.25 (#2) · ⚡ 9.6%
  13. " Mark" recently introduced sparsest Markov representation (SM↕ -25.50 (#2) · ⚡ 8.5%
  14. " as", also referred to as Bayesian networks which are↕ -23.75 (#2) · ⚡ 8.0%
  15. " maximum" model that entails the maximum number of d-se↕ -21.75 (#1) · ⚡ 7.4%
  16. "-time" is sufficient for polynomial-time algorithms [@G↕ -21.62 (#2) · ⚡ 7.2%
  17. ":**"------↵↵**Keywords:** Directed graphical Models,↕ -21.38 (#1) · ⚡ 9.1%
  18. "ifi" the issue of identifiability for general directed↕ -21.00 (#2) · ⚡ 7.0%
  19. " " them have been discussed [@Zhang2↕ -21.00 (#1) · ⚡ 8.6%
  20. " the"$G$ given the data-generating distribution↕ -20.12 (#1) · ⚡ 5.2%
Qwen 3.5 9B Base · L26/C68

Suffix Sh-

Confidence 5/5
Component profile — C68 · layer 26 · negative tail

Logit-lens tokens · negative

#1"(sh"#2"akespeare"#3"neider"#4"apiro"#5".rmtree"#6"ADOW"#7"ifter"#8"ields"#9"ifting"#10"reiber"

R-lens tokens · negative

#1"(sh"#2"/Sh"#3"neider"#4"ADOW"#5"/sh"#6".rmtree"#7"uffling"#8"akespeare"#9"gun"#10"apesh"

Skewness

-12.89

Excess kurtosis

330.72rank #106

Logcosh deviation

0.1842rank #69

Top-score occurrences · negative

  1. "Sh" some philosophy.↵↵Sheldon was Born and↕ -48.00 (#1) · ⚡ 37.5%
  2. "Sh"us monkey.↵↵Shroud, Rhône Glacier↕ -48.00 (#1) · ⚡ 41.2%
  3. " Sh" ride.↵↵The Shogunate’s life↕ -44.75 (#1) · ⚡ 31.6%
  4. "Sh" Kumar added.↵↵Shahin Noore↕ -44.50 (#1) · ⚡ 36.9%
  5. "Sh" making a very "Shayna" face↕ -44.50 (#1) · ⚡ 38.1%
  6. "Sh" apocalypse is imminent↵↵Shlomi Dole↕ -44.25 (#1) · ⚡ 35.0%
  7. " Sh" group's president, Shalesh Kumb↕ -43.50 (#1) · ⚡ 37.9%
  8. "Sh"56]).↵↵Shells of bival↕ -42.75 (#1) · ⚡ 29.1%
  9. " Sh"Resources > Materials and Shaders | Resources |↕ -42.25 (#1) · ⚡ 35.4%
  10. "Sh" like this." "Shh, shh↕ -42.25 (#1) · ⚡ 33.8%
  11. "Sh" de su expansión (Shutterstock)↵↵La↕ -41.25 (#1) · ⚡ 34.6%
  12. " Sh"*648 Sherryl Snod↕ -41.00 (#1) · ⚡ 31.6%
  13. " Sh" make-up experience, Shu Uemura↕ -40.75 (#1) · ⚡ 24.5%
  14. " Sh"↵For example, Shuichi Kurabay↕ -40.50 (#1) · ⚡ 32.4%
  15. " Sh" basis↵Hindu Shaktism↵the↕ -39.75 (#1) · ⚡ 21.9%
  16. " sh" uncover and expose the shenanigans of the↕ -39.75 (#1) · ⚡ 36.1%
  17. "sh", and give↵shippers an edge over↕ -39.50 (#1) · ⚡ 31.8%
  18. " sh"↵↵It's a shrewd move,↕ -39.50 (#1) · ⚡ 33.4%
  19. "sh" changed, with “sharper oversight and↕ -39.25 (#1) · ⚡ 34.4%
  20. " Sh" victory over a tame Shrewsbury in↕ -39.25 (#1) · ⚡ 32.2%
Qwen 3.5 9B Base · L26/C1210

Deep-learning framework/model

Confidence 5/5
Component profile — C1210 · layer 26 · positive tail

Logit-lens tokens · positive

#1" Tensor"#2" trainable"#3" softmax"#4" keras"#5"Tensor"#6" TensorFlow"#7" tensor"#8"训练"#9" LSTM"#10" tensorflow"

R-lens tokens · positive

#1" Tensor"#2" trainable"#3"Tensor"#4" keras"#5" TensorFlow"#6"训练"#7" tensor"#8" tensorflow"#9" training"#10" softmax"

Skewness

+4.81

Excess kurtosis

95.20rank #1085

Logcosh deviation

0.0793rank #1211

Top-score occurrences · positive

  1. " the"↵This includes both the weights and the architecture↕ +38.25 (#1) · ⚡ 22.9%
  2. " the" both the weights and the architecture.↵↵Models↕ +38.25 (#1) · ⚡ 22.9%
  3. " both"().↵This includes both the weights and the↕ +36.75 (#1) · ⚡ 22.0%
  4. " the"↵ones needed for the model.↵↵You↕ +36.00 (#1) · ⚡ 23.0%
  5. "(\""Error: Tensor Tensor("Sigmoid_2↕ +34.75 (#1) · ⚡ 15.3%
  6. "\n", or only the↵ones needed for the↕ +33.25 (#1) · ⚡ 19.2%
  7. "}."('model{0}.h5'.format↕ +31.25 (#1) · ⚡ 10.3%
  8. " all" decide to save out all the variables, or↕ +31.25 (#1) · ⚡ 16.8%
  9. " ("10 neurons each (size ~50↕ +30.88 (#1) · ⚡ 16.5%
  10. " with":↵↵I tried with K.clear_session(),↕ +30.25 (#1) · ⚡ 20.1%
  11. " during" the main graph) during writing the graph summary↕ +29.88 (#1) · ⚡ 12.9%
  12. " a" keras↵↵I have a set of Keras↕ +29.75 (#1) · ⚡ 18.3%
  13. " \""n",↵ "layers = [\n↕ +29.38 (#1) · ⚡ 19.1%
  14. "igmoid" Tensor Tensor("Sigmoid_2:0↕ +29.00 (#1) · ⚡ 9.4%
  15. " the" graph) during writing the graph summary?↵↵↕ +28.50 (#1) · ⚡ 9.0%
  16. " keras" slow model load with keras↵↵I have a↕ +28.50 (#1) · ⚡ 13.6%
  17. " model" other, what are model variables -- do you↕ +28.12 (#1) · ⚡ 15.1%
  18. " each" 10 neurons each (size ~5↕ +27.88 (#1) · ⚡ 15.3%
  19. " \""\n",↵ "data_in_shape =↕ +27.50 (#1) · ⚡ 15.8%
  20. " network" for l in network.layers:↵ ↕ +27.25 (#1) · ⚡ 9.4%
Qwen 3.5 9B Base · L28/C3025

Suffix Var-

Confidence 4/5
Component profile — C3025 · layer 28 · negative tail

Logit-lens tokens · negative

#1"iances"#2"ly"#3"ieg"#4"iations"#5"nish"#6"iance"#7"ials"#8"iazione"#9"iable"#10"ious"

R-lens tokens · negative

#1"iances"#2"ly"#3"ieg"#4"iations"#5"iance"#6"nish"#7"ials"#8"iable"#9"ious"#10"iazione"

Skewness

-1.94

Excess kurtosis

33.56rank #2240

Logcosh deviation

0.0407rank #3026

Top-score occurrences · negative

  1. " Var"uforums. Varunendra is very↕ -35.25 (#1) · ⚡ 19.0%
  2. " var"A fuse and a varistor of type S↕ -33.25 (#1) · ⚡ 15.7%
  3. " Vari" and control samples. Variations in guanase↕ -31.38 (#1) · ⚡ 16.5%
  4. " var"↵↵Use the PHP var_export() function as↕ -29.50 (#1) · ⚡ 13.8%
  5. " var" first selection criteria in varietal improvement programs↕ -29.12 (#1) · ⚡ 14.4%
  6. ",var",lam,pen,varargin) where↕ -28.75 (#1) · ⚡ 15.7%
  7. " var" fine art usage picture varnishes should be↕ -28.62 (#1) · ⚡ 12.2%
  8. " var"↵↵Passing a var from one Extended function↕ -28.12 (#1) · ⚡ 19.6%
  9. "(var"n(mean, sqrt(variance)); // Assuming↕ -27.75 (#1) · ⚡ 13.7%
  10. " var" but as receipt of varicella vaccine at↕ -27.25 (#1) · ⚡ 16.5%
  11. " Var" $Y$ enjoying Varolin’s [*density↕ -26.75 (#1) · ⚡ 12.2%
  12. "var"071 /var/run/acpid.socket↕ -26.38 (#1) · ⚡ 10.2%
  13. " Var"0 years. Mr Varadkar approached Mr↕ -26.25 (#1) · ⚡ 15.1%
  14. " var" phlebectomy of varices were performed under↕ -26.12 (#1) · ⚡ 10.5%
  15. " var" the future. Our varnishes are labelled↕ -25.75 (#1) · ⚡ 12.6%
  16. " vari" and Ingard's variational-based treatment of↕ -25.12 (#1) · ⚡ 10.7%
  17. "var" expected: "/var/run/docker.sock",↕ -24.75 (#1) · ⚡ 9.5%
  18. " var" - A clear gloss varnish which provides permanent↕ -24.62 (#1) · ⚡ 12.9%
  19. "var"alert -a /var/log/audit/↕ -23.88 (#1) · ⚡ 9.0%
  20. " Var" Taoiseach Leo Varadkar, it↕ -23.75 (#1) · ⚡ 11.6%
Qwen 3.5 9B Base · L30/C2683

Expecting "see"

Confidence 5/5
Component profile — C2683 · layer 30 · positive tail

Logit-lens tokens · positive

#1" see"#2"见"#3"see"#4" See"#5"See"#6" sees"#7" seeing"#8" seen"#9"見"#10"见的"

R-lens tokens · positive

#1" see"#2"见"#3"see"#4" See"#5"See"#6" sees"#7" seeing"#8" seen"#9"見"#10"见的"

Skewness

+1.54

Excess kurtosis

13.75rank #3380

Logcosh deviation

0.0437rank #2684

Top-score occurrences · positive

  1. " to" Would you have liked to have seen them struggle↕ +22.62 (#1) · ⚡ 9.6%
  2. " be" as evidence remains to be seen at trial.↕ +20.50 (#1) · ⚡ 7.3%
  3. " to", I’d like to see something a little↕ +20.38 (#1) · ⚡ 10.0%
  4. " be" crystal display panel can be seen from the outside↕ +19.25 (#1) · ⚡ 8.0%
  5. " to" good idea to arrange to see your GP or↕ +19.00 (#1) · ⚡ 7.1%
  6. " we" second, and what we see in the jQuery↕ +18.25 (#1) · ⚡ 8.4%
  7. " like"4, I’d like to see something a↕ +18.12 (#1) · ⚡ 6.8%
  8. " to" thought you might like to see the Baby Boy↕ +17.75 (#1) · ⚡ 7.5%
  9. " can".↵↵So you can see, metals,↕ +17.62 (#2) · ⚡ 6.1%
  10. " to" up. They want to see that; they↕ +17.62 (#1) · ⚡ 7.7%
  11. "_to"current, :allowed_to?)↵ .↕ +17.62 (#2) · ⚡ 5.6%
  12. " be"____↵As can be estimated from the values↕ +17.50 (#1) · ⚡ 4.1%
  13. " to" user is not allowed to see watchers but there↕ +17.12 (#1) · ⚡ 9.3%
  14. " can" lock↵As you can see on the table↕ +16.88 (#1) · ⚡ 5.5%
  15. " to" it’s not hard to see why.↵↵↕ +16.62 (#1) · ⚡ 5.2%
  16. " to" that others may prefer to see children in these↕ +16.38 (#1) · ⚡ 7.6%
  17. "iscopal" It became an episcopal see under Gregory VII↕ +16.12 (#2) · ⚡ 3.3%
  18. "." either express or implied. See the↵ *↕ +16.12 (#1) · ⚡ 6.6%
  19. " we" hairless cat when we see one. Although↕ +16.00 (#2) · ⚡ 5.6%
  20. " can" spring, as you can tell from the dens↕ +16.00 (#2) · ⚡ 6.3%