Neuron Explorer

Qwen3-0.6B

the spot before a year

A neuron in layer 9 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the blank spot just before a year, such as the '1970' in 'Occupational Safety and Health Act of 1970'. The opposite side fires on single digits inside number ranges. Turning it up made the text wander without a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    and preventing exposure in the workplace, see OSHA's Carbon Monoxide Poisoning Quick Cards (in English and Spanish).⏎Under the Occupational Safety and Health Act of 1970, employers are responsible for providing safe and healthful workplaces for

  2. #2 peak

    for no other reason than gender. Now is the time to change that.⏎The Paycheck Fairness Act, a long overdue amendment to the Equal Pay Act of 1963, (S. 182) was introduced by

  3. #3 peak

    this extreme idea for enforcement was abandoned.⏎This present day federal building code for our homes came out of “The National Manufactured Housing Construction and Safety Standards Act of 1974"and became federal law on June 15,

  4. #4 peak

    health standards, encourages voluntary compliance through consultation and partnerships, and promotes safety training and education for workers and employers. Since Congress passed the Occupational Safety and Health Act of 1970 (OSH Act), which created OSHA as an agency within

  5. #5 peak

    have advanced. Now more than ever, educators and students are being held accountable for mastery of these subjects. With the institution of the No Child Left Behind Act of 2001 and the high school exit examinations in use in at least

  6. #6 peak

    0597 or make an appointment online at www.MyFortMyersDentist.com.<|endoftext|>In a world where 1 in 7 people are displaced, your kindness is just condescension.⏎The piro

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: the 1000000000000000000000000000000000000000000000

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    Numbers and Operations in Base Ten Task Cards (NBT) 14-18 Operations and Algebraic Thinking Task Cards (OA) 19-23 Geometry Task Cards (G) 24-28 Measurement and Data

  2. #2 peak

    Christmas season, make ornaments that will help your kids consider the biblical story of the coming of Jesus.⏎The Star of Bethlehem⏎Read Matthew 2:1-12 with your kids.⏎Before making your ornaments, discuss:⏎- Where was Jesus

  3. #3 peak

    Reviews.⏎The results indicated that improving lifestyle habits among children exposed to gestational diabetes can prevent poor adiposity values.⏎The study included children aged 2 to 14 years, 105 of whom had been exposed to gestational diabetes

  4. #4 peak

    no control. A quick check of Jesus’ parables shows the relationships between slaves/servants and wealthy people. (See Matthew 20:1-16, Luke 16:1-13 or Luke 19

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L8:1165 excites
  2. L8:119 suppresses
  3. L8:162 suppresses
  4. L8:394 suppresses
  5. L8:283 suppresses
  6. L8:421 excites
  7. L4:294 suppresses
  8. L8:457 excites year digits in copyright and dates
  9. L7:2878 excites
  10. L6:346 suppresses
  11. L6:371 suppresses
  12. L6:3042 excites
  13. L8:465 suppresses
  14. L8:240 suppresses
  15. L3:1949 excites
  16. L8:2783 suppresses
  17. L7:6 suppresses
  18. L6:91 weakly excites
  19. L6:2086 weakly excites
  20. L8:2052 weakly suppresses
  21. L8:350 weakly suppresses blank space before a year
  22. L8:1321 weakly suppresses
  23. L8:320 weakly excites
  24. L6:184 weakly excites

Feeds later layers

  1. L10:152 strongly suppresses
  2. L10:121 suppresses
  3. L15:42 suppresses
  4. L14:542 suppresses
  5. L10:220 suppresses hyphen inside compound words
  6. L12:92 suppresses
  7. L13:1342 suppresses digits inside years like 2017
  8. L12:2434 excites
  9. L11:577 suppresses
  10. L11:329 suppresses number or citation fragment split
  11. L15:368 suppresses
  12. L16:256 suppresses number right after 'age of' or 'ages'
  13. L13:999 suppresses
  14. L14:759 suppresses
  15. L13:866 suppresses
  16. L10:232 suppresses
  17. L10:42 suppresses
  18. L15:168 suppresses
  19. L15:1601 suppresses
  20. L11:148 suppresses
  21. L10:40 excites
  22. L15:376 excites
  23. L10:184 excites
  24. L11:300 suppresses

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -7.93 and -7.631 values between -6.43 and -6.132 values between -5.83 and -5.532 values between -5.23 and -4.932 values between -4.93 and -4.633 values between -4.63 and -4.324 values between -4.32 and -4.021 values between -4.02 and -3.723 values between -3.72 and -3.424 values between -3.42 and -3.1210 values between -3.12 and -2.825 values between -2.82 and -2.521 values between -2.52 and -2.228 values between -2.22 and -1.925 values between -1.92 and -1.625 values between -1.62 and -1.329 values between -1.32 and -1.0211 values between -1.02 and -0.727 values between -0.72 and -0.41396 values between -0.41 and -0.112972 values between -0.11 and 0.19383 values between 0.19 and 0.4939 values between 0.49 and 0.7916 values between 0.79 and 1.096 values between 1.09 and 1.393 values between 1.39 and 1.693 values between 1.69 and 1.993 values between 1.99 and 2.291 values between 3.20 and 3.501 values between 3.50 and 3.801 values between 3.80 and 4.10offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. essential
  2. standen
  3. ashing
  4. earing
  5. stead
  6. 데
  7. uten
  8. YE
  9. Broken
  10. QUE

Suppresses

  1. 良心
  2. 镧
  3. acies
  4. ivy
  5. oo
  6. o
  7. 心
  8. 的房子
  9. 喉
  10. ò