Neuron Explorer

Qwen3-0.6B

document boundary, code keywords next

A neuron in layer 27 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 marker that ends one document and starts another. When it is turned up, the model tends to write programming words like 'class', 'static' and 'public'.

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

    old<|endoftext|>Immune cells in the brain have surprising influence on sexual behavior⏎Relatively uncommon

  2. #2 peak

    data from TC1.<|endoftext|>The concept of welfare state⏎We all know that most of the modern democratic countries

  3. #3 peak

    locate.<|endoftext|>Science Fair Project Encyclopedia⏎The World Factbook⏎The World Factbook is an

  4. #4 peak

    needed around planes en route to keep them safely separated.<|endoftext|>FEDERAL Hill House is a squat building in central Providence, within earshot

  5. #5 peak

    /definition/acanthosis nigricans>.<|endoftext|>Dronacharya was the teacher of both the Kauravas and Pand

  6. #6 peak

    to like.<|endoftext|>The late Dominica-born writer Jean Rhys, best known for her novel “

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: perhaps you may be able to avoid criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism criticism

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

    ’ Google), you’re sure to spot subtle references that talk about ‘timeless content’.⏎No, this doesn’t necessarily mean that you have to whip up a Silmarillion every time you open MS Word – all you need to do is

  2. #2 peak

    a strand.⏎Very few organic molecules can with stand high temperatures for very⏎long. Heat provides the activation energy to break covalent bonds and⏎denature/destroy molecular structure. Proteins are particularly sensitive.⏎There are exceptions, as in

  3. #3 peak

    , that active or hyperactive kids moving in rhythm with the group, such as in walking meditation, derive a calming effect. 11. Meditation is non-competitive, and each can participate within the limits of their own ability, with adaptations

  4. #4 peak

    . Instances of the class Class represent classes and interfaces in a running Java application. If we do not explicitly write a constructor for a class, the Java compiler builds a default constructor for that class. so now we can use this to create objects.

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. L19:542 suppresses
  2. L26:120 suppresses
  3. L22:633 suppresses
  4. L23:1647 excites
  5. L22:39 suppresses
  6. L14:67 suppresses
  7. L18:284 suppresses
  8. L15:524 excites
  9. L17:2265 excites
  10. L19:1372 suppresses
  11. L21:155 suppresses
  12. L24:721 suppresses
  13. L20:205 excites
  14. L26:399 suppresses
  15. L26:612 excites end-of-text boundary tokens
  16. L17:95 suppresses
  17. L14:1636 excites
  18. L23:81 excites
  19. L22:1816 weakly excites
  20. L10:38 weakly suppresses
  21. L9:18 weakly suppresses text boundary after a document break
  22. L21:162 weakly excites
  23. L21:131 weakly excites
  24. L17:35 weakly excites

Feeds later layers

None: this is the last layer.

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.

2 values between -6.41 and -5.462 values between -5.46 and -4.516 values between -4.51 and -3.5511 values between -3.55 and -2.6051 values between -2.60 and -1.65391 values between -1.65 and -0.702661 values between -0.70 and 0.26688 values between 0.26 and 1.2135 values between 1.21 and 2.1616 values between 2.16 and 3.1112 values between 3.11 and 4.0710 values between 4.07 and 5.024 values between 5.02 and 5.973 values between 5.97 and 6.925 values between 6.92 and 7.882 values between 7.88 and 8.831 values between 8.83 and 9.782 values between 9.78 and 10.733 values between 10.73 and 11.691 values between 11.69 and 12.641 values between 16.45 and 17.401 values between 30.74 and 31.69offnegativepositiveoften

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. class
  2. static
  3. static
  4. class
  5. public
  6. public
  7. struct
  8. function
  9. template
  10. auto

Suppresses

  1. ее
  2. ierrez
  3. вен
  4. ämpfe
  5. 閣
  6. -care
  7. 회의
  8. criticism
  9. ậm
  10. unately