Neuron Explorer

Qwen3-0.6B

unclear: verbs after subject, line breaks

A neuron in layer 17 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 responds to a mixed set of spots, including verbs right after a subject ('I am', 'It is', 'You are') and line breaks after a heading or question. The negative side is clearer: it dips on the colon or line break that follows a word like 'Excerpt'. Turning it up or down gave loose, repetitive text, so the function is not clear.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    they all sitting? Can you find all the possible ways?⏎What is the best way to shunt these carriages so that each train⏎can continue its journey?⏎If you split the square into these

  2. #2 peak

    ini Araghi, BSc, MPH, of the University of Birmingham, and colleagues concluded.⏎Disclosure: The researchers report no relevant financial disclosures.<|endoftext|>You are here⏎Edward Elgar (1857-1934)

  3. #3 peak

    Yersiniae cause severe intestinal infections. Studies using Yersinia pseudotuberculosis as a model organism aim to elucidate the infection mechanisms of these...⏎Researchers from the

  4. #4 peak

    I am pleased to announce that gold processing operations have⏎Ore processing operations at Newmont

  5. #5 peak

    I am concerned about the health impacts of mould. What should I do?⏎If you are

  6. #6 peak

    joined to form a⏎quadrilateral. What do you notice about this quadrilateral as the⏎dynamic image changes? Can you prove your conjecture?⏎It is impossible to trisect an angle using

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: else, but, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else, else,

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

    The foxes have holes, and the birds of the heaven have nests; but the Son of man hath not where to lay his head.⏎Young's Literal Translation⏎and Jesus said to him, 'The foxes have holes, and the f

  2. #2 peak

    ⏎- Subject: Drama - World⏎- Type: Essay⏎- Paper: #87743533⏎Excerpt from Essay :⏎Perceptions of Che Guevera⏎PERCEPTIONS OF CHE GUE

  3. #3 peak

    he traces the history of instant communication from ancient times to today.⏎→ Event: Tom Standage | Barnes and Noble Wednesday 6pm | Information⏎Excerpt: Writing on the Wall⏎INTRODUCTION: CICERO’S WEB⏎Not to

  4. #4 peak

    off from their mouth.⏎Jeremiah 22:21 "I spoke to you in your<|endoftext|>How’s your facility looking this flu season?

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. L15:975 suppresses
  2. L10:931 suppresses
  3. L14:812 suppresses
  4. L4:402 weakly suppresses
  5. L9:257 weakly suppresses
  6. L6:1082 weakly suppresses
  7. L4:653 weakly suppresses "that" after "such" or "such a way"
  8. L9:105 weakly suppresses
  9. L14:2017 weakly suppresses
  10. L15:83 weakly suppresses
  11. L14:1282 weakly suppresses
  12. L15:679 weakly suppresses
  13. L8:686 weakly excites "of" after full text or version
  14. L3:944 weakly excites
  15. L16:1662 weakly excites
  16. L1:2094 weakly excites
  17. L16:2743 weakly suppresses
  18. L16:1080 weakly suppresses
  19. L15:2930 weakly suppresses
  20. L16:555 weakly suppresses
  21. L13:114 weakly suppresses
  22. L16:1086 weakly suppresses
  23. L16:2599 weakly suppresses
  24. L16:1652 weakly excites

Feeds later layers

  1. L23:221 suppresses
  2. L18:769 suppresses
  3. L27:1006 suppresses
  4. L21:478 suppresses
  5. L18:928 suppresses
  6. L19:1593 suppresses
  7. L27:568 suppresses
  8. L19:2294 suppresses
  9. L20:2429 suppresses
  10. L18:1775 suppresses
  11. L25:476 suppresses comma before 'but' in contrasts
  12. L19:934 weakly suppresses
  13. L19:2997 weakly suppresses
  14. L19:3030 weakly suppresses
  15. L18:2066 weakly suppresses
  16. L27:2076 weakly suppresses
  17. L25:1425 weakly suppresses
  18. L19:1930 weakly suppresses
  19. L18:2471 weakly excites
  20. L26:1232 weakly suppresses
  21. L19:2590 weakly suppresses
  22. L18:2007 weakly suppresses
  23. L23:621 weakly suppresses
  24. L27:1750 weakly 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 -6.36 and -6.121 values between -5.40 and -5.161 values between -4.68 and -4.441 values between -4.44 and -4.201 values between -3.24 and -3.002 values between -3.00 and -2.773 values between -2.77 and -2.531 values between -2.53 and -2.294 values between -2.29 and -2.055 values between -2.05 and -1.815 values between -1.81 and -1.577 values between -1.57 and -1.3320 values between -1.33 and -1.0924 values between -1.09 and -0.8530 values between -0.85 and -0.6178 values between -0.61 and -0.37410 values between -0.37 and -0.131788 values between -0.13 and 0.111105 values between 0.11 and 0.35342 values between 0.35 and 0.5954 values between 0.59 and 0.8313 values between 0.83 and 1.073 values between 1.07 and 1.303 values between 1.30 and 1.541 values between 1.54 and 1.783 values between 1.78 and 2.021 values between 2.74 and 2.981 values between 2.98 and 3.22offnegativepositiveoften

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. og
  2. umbs
  3. kov
  4. ogg
  5. aw
  6. olo
  7. oda
  8. ol
  9. osa
  10. leh

Suppresses

  1. then
  2. 但也
  3. then
  4. 但它
  5. 然后
  6. Then
  7. Then
  8. 但实际上
  9. 否则
  10. 也因此