Neuron Explorer

Qwen3-0.6B

numbers after 'y =' or 'cosine of'

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a specific number that comes right after a word like 'y =' or 'cosine of', as in 'y = 2x' or 'cosine of 5x'. Turning it up made the model's output drift into long, rambling word salad, and it did not produce 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

    how to identify parallel and perpendicular lines.⏎Let's take some examples to understand this.⏎Considering two equations:⏎y = 2x +3 and y = 2x+5⏎On comparing y = 2x +3 and y =

  2. #2 peak

    have our du.⏎- Our du is 5 -- let me circle that and let me do that in that blue color --⏎- is 5 cosine of 5x dx. So we can rewrite this entire expression as --⏎

  3. #3 peak

    dx--⏎- sorry, I need cosine of 5x dx, just like that. So when you look over here,⏎- you have a cosine of 5x dx,

  4. #4 peak

    00000000 IP: 10101100.0001000x.xxxxxxxx.xxxxxxxx Each subnet has 2500

  5. #5 peak

    00 n = 15 NETWORK IP: 10101100.0001000x.xxxxxxxx.xxxxxxxx SUBNET IP: 10101100.0

  6. #6 peak

    we know that,⏎Lines with the same slope are parallel.⏎Hence, the given lines are parallel.⏎Let's consider another set of equations.⏎y = 2x +3 and 4y = –2x+10⏎On

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: cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive

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

    middle school students.19<|endoftext|>毋忘在莒(Pinyin: Wu Wang Zai Ju) is a famous Chinese idiom (成語, Pinyin: Cheng Yu). It is sort of like the Latin idiom left in the European

  2. #2 peak

    legal system, early Confucian scholars emphasized the concepts of morality and ritualism. The term “ritual propriety” (禮, pinyin: lǐ), describes the “proper” social relationships and the set of rituals which regulate them

  3. #3 peak

    We hold these Truths to be self-evident, that all Men are created equal, that they are endowed by their

  4. #4 peak

    , Seobeol). In 503, King Jijeung standardized on the characters 新羅(신라), which in Modern Korean is pronounced "Shilla

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. L24:495 strongly suppresses
  2. L23:1915 excites
  3. L23:1848 excites
  4. L23:2224 excites
  5. L22:918 suppresses
  6. L23:1668 suppresses
  7. L22:411 suppresses
  8. L24:2291 suppresses
  9. L24:1119 suppresses
  10. L24:2406 suppresses
  11. L20:2900 excites
  12. L23:387 suppresses
  13. L23:1128 excites
  14. L21:972 excites
  15. L22:859 suppresses
  16. L10:1510 excites
  17. L23:2862 excites
  18. L23:1764 excites
  19. L22:1121 suppresses
  20. L23:2997 weakly suppresses
  21. L21:2318 weakly suppresses
  22. L3:2159 weakly excites
  23. L20:2955 weakly excites
  24. L20:1450 weakly suppresses

Feeds later layers

  1. L27:560 strongly suppresses
  2. L27:1128 strongly suppresses
  3. L27:589 strongly suppresses
  4. L27:534 strongly suppresses
  5. L26:1213 suppresses
  6. L26:957 suppresses
  7. L27:768 weakly suppresses
  8. L26:1038 weakly excites
  9. L27:1147 weakly suppresses
  10. L26:293 weakly excites
  11. L26:1278 weakly suppresses
  12. L27:1441 weakly suppresses
  13. L26:162 weakly excites
  14. L26:2495 weakly suppresses
  15. L27:566 weakly suppresses
  16. L27:2795 weakly suppresses
  17. L27:1080 weakly excites
  18. L27:1517 weakly suppresses
  19. L26:2784 weakly suppresses
  20. L26:1327 weakly excites opening parenthesis before an acronym
  21. L27:2858 weakly suppresses
  22. L27:2031 weakly suppresses
  23. L27:89 weakly excites
  24. L27:2199 weakly excites

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 -12.95 and -12.352 values between -9.34 and -8.741 values between -8.14 and -7.541 values between -5.73 and -5.132 values between -5.13 and -4.532 values between -4.53 and -3.923 values between -3.92 and -3.325 values between -3.32 and -2.7212 values between -2.72 and -2.1218 values between -2.12 and -1.5239 values between -1.52 and -0.91196 values between -0.91 and -0.312464 values between -0.31 and 0.291082 values between 0.29 and 0.8963 values between 0.89 and 1.495 values between 1.49 and 2.106 values between 2.10 and 2.701 values between 2.70 and 3.302 values between 3.30 and 3.901 values between 3.90 and 4.501 values between 6.31 and 6.911 values between 10.52 and 11.13offnegativepositiveoften

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. @register
  2. maids
  3. ugin
  4. semblies
  5. Posting
  6. erner
  7. 渤
  8. 楣
  9. esign
  10. ificant

Suppresses

  1. X
  2. X
  3. x
  4. >X
  5. x
  6. ,X
  7. .X
  8. _X
  9. -X
  10. →X