Neuron Explorer

Qwen3-0.6B

unclear: weak, scattered tokens

A neuron in layer 13 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 only weakly, and to a scattered set of word pieces with no obvious shared meaning. Its strongest examples are hard to tell apart from ordinary text, so it is best treated as a mystery.

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

    The answer to soaring gas prices lies in renewable heating, a technology that is ready to replace gas

  2. #2 peak

    moral principle)⏎3. We ought to do something to help them. (ought)⏎

  3. #3 peak

    Copyright © 1997, by Random House, Inc., on Infoplease.<|endoftext|>This was

  4. #4 peak

    Pretty Balloons (counting/number sentences) (Resource Book Only) eBookKindergarten|Grade 1⏎Please Note: This ebook is a digital download, NOT a physical product. After purchase, you

  5. #5 peak

    Temp. (Celsius)||-65|⏎|Max. Temp. (Celsius)||165|⏎There are no reviews yet.<|endoftext|>Talk about doing something in the blink of an eye.⏎Google wants to take that phrase to a

  6. #6 peak

    - In search of their lost ones.<|endoftext|>The Verbal Reasoning section of the Graduate

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: I have always been a fan of the book, but I have never been a fan of the book, but I have always liked the book, but I have a history of liking the book, but I have a mix of like and dislike

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

    aton & Olivier Dupriez, 2011.⏎"Purchasing Power Parity Exchange Rates for the Global Poor,"⏎American Economic Journal: Applied Economics,⏎American Economic Association, vol. 3(2), pages 137

  2. #2 peak

    spell”November 23rd, 2010 - 6:33 pm ICT by ANI⏎London, Nov 23 (ANI): A study has suggested that Internet chatrooms and social networking sites are to be

  3. #3 peak

    aceous amberJune 16th, 2010 - 1:53 pm ICT by ANI⏎London, June 16 (ANI): Palaeontologists have discovered two mammal hairs encased in amber that

  4. #4 peak

    . Dwapara Yug<|endoftext|>Simple technique appears to be safe and effective, review suggests⏎By Robert Preidt⏎MONDAY, Oct. 15 (HealthDay News) -- A technique called the "mother's kiss" is a

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. L10:333 excites
  2. L11:171 excites
  3. L10:124 suppresses
  4. L10:1578 excites
  5. L12:1501 suppresses
  6. L5:780 suppresses
  7. L12:3045 suppresses
  8. L9:716 excites
  9. L10:3055 suppresses
  10. L10:169 excites
  11. L12:1979 excites
  12. L12:1115 excites
  13. L12:2174 suppresses mixed, web boilerplate phrases
  14. L10:812 excites
  15. L9:2521 suppresses
  16. L5:986 suppresses
  17. L11:17 suppresses
  18. L12:2036 excites
  19. L9:2085 excites
  20. L10:1526 excites
  21. L8:957 excites
  22. L9:2622 excites
  23. L11:29 excites
  24. L10:2973 suppresses

Feeds later layers

  1. L14:1226 strongly excites
  2. L14:267 strongly excites
  3. L17:892 strongly excites
  4. L16:200 suppresses
  5. L15:2628 excites
  6. L14:2941 suppresses
  7. L15:23 suppresses
  8. L14:2974 suppresses
  9. L15:1178 excites
  10. L14:932 suppresses
  11. L20:543 excites
  12. L14:84 excites
  13. L16:1163 excites
  14. L14:2556 excites
  15. L16:883 excites
  16. L16:248 suppresses
  17. L19:417 excites
  18. L17:722 excites
  19. L22:157 excites
  20. L17:661 excites names in website boilerplate
  21. L16:1543 excites
  22. L25:388 excites
  23. L20:38 suppresses
  24. L27:581 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 -0.71 and -0.681 values between -0.68 and -0.653 values between -0.65 and -0.613 values between -0.61 and -0.585 values between -0.58 and -0.556 values between -0.55 and -0.517 values between -0.51 and -0.489 values between -0.48 and -0.4416 values between -0.44 and -0.4137 values between -0.41 and -0.3831 values between -0.38 and -0.3455 values between -0.34 and -0.3167 values between -0.31 and -0.27105 values between -0.27 and -0.2498 values between -0.24 and -0.21140 values between -0.21 and -0.17192 values between -0.17 and -0.14227 values between -0.14 and -0.11246 values between -0.11 and -0.07316 values between -0.07 and -0.04353 values between -0.04 and -0.00383 values between -0.00 and 0.03345 values between 0.03 and 0.06296 values between 0.06 and 0.10220 values between 0.10 and 0.13185 values between 0.13 and 0.17171 values between 0.17 and 0.20124 values between 0.20 and 0.2386 values between 0.23 and 0.2756 values between 0.27 and 0.3039 values between 0.30 and 0.3336 values between 0.33 and 0.3715 values between 0.37 and 0.4010 values between 0.40 and 0.447 values between 0.44 and 0.473 values between 0.47 and 0.503 values between 0.50 and 0.545 values between 0.54 and 0.572 values between 0.57 and 0.604 values between 0.60 and 0.64offnegativepositiveoften

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. .should
  2. irse
  3. ENTER
  4. Unidos
  5. tro
  6. roph
  7. '||
  8. indexed
  9. Harlem
  10. paces

Suppresses

  1. odal
  2. https
  3. apter
  4. ита
  5. antom
  6. ạ
  7. latex
  8. angle
  9. airy
  10. odb