Neuron Explorer

Qwen3-0.6B

unclear: 'decre' and split word fragments

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires most strongly on the start of the word 'decree' (written as 'decre' before its ending) and on a few other word pieces in historical or legal text. Turning it up made the model wander off into unrelated writing.

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

    sentences. The UN Working Group on Enforced or Involuntary Disappearances (WGEID) stated that while Chile has

  2. #2 peak

    only way to ensure that the knowledge that leads to long life, and peace and prosperity in a nation is passed to each generation. “These are the commands, decrees and laws the Lord your God directed me to teach you to observe in the

  3. #3 peak

    870)|⏎|Convoked by||Paul III|⏎about 255 during the final sessions⏎Documents and statements⏎Seventeen dogmatic decrees covering then-disputed aspects of Catholic religion⏎|Chronological list of ec

  4. #4 peak

    his city's role in the Trans-Atlantic slave trade. He acknowledged that London was "tainted" by it<|endoftext|>Nearly three months after the Treaty of Paris was signed ending the American Revolution, the last British soldiers withdraw from New York City,

  5. #5 peak

    Poles turned from the Roman Church, winning a measure of religious freedom. Some church buildings were put to Protestant use. Laws forbade the execution of Roman Church decrees

  6. #6 peak

    housing, health and educational needs as well as economic, social and cultural rights.⏎The United Nations Convention on the Rights of the Child, to which New Zealand is a signatory, says that social services is also about ensuring that in all that we

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: what if the *red* *and* *red* *and* *red* *and* *red* *and* *red* *and* *red* *and* *red* *and* *red*

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

    loss of enthusiasm.⏎33. Failure will never overtake me if my determination to succeed is strong enough.⏎34. Failure doesn

  2. #2 peak

    one by one, they find themselves overwhelmed by guilt and contrition over their support of a murderous regime. He shows how their awakened moral reckonings and higher interests overrode lifetime habits and disciplines on the road to “the wolf’s lair

  3. #3 peak

    place on the American streets. It never belonged to the streets like during former days of jazz. This fusion of music was styled in a positive manner that made it overtake rap in early ‘90s (Keyes 245).

  4. #4 peak

    -bedded poorly sorted sand and gravel. The aquifer is composed of hydraulically connected alluvial and terrace deposits that unconformably overlie the Permian-age Formations.⏎The aquifer boundaries are from a ground

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. L20:1074 weakly excites
  2. L3:1101 weakly suppresses
  3. L24:370 weakly suppresses
  4. L0:2774 weakly suppresses
  5. L9:833 weakly excites
  6. L19:758 weakly suppresses
  7. L15:930 weakly suppresses
  8. L21:2818 weakly excites
  9. L5:1838 weakly excites
  10. L24:1551 weakly excites
  11. L14:655 weakly suppresses
  12. L23:2421 weakly excites
  13. L19:1723 weakly excites
  14. L10:2797 weakly suppresses
  15. L21:1364 weakly excites
  16. L20:357 weakly suppresses
  17. L9:2970 weakly suppresses
  18. L14:2228 weakly suppresses
  19. L23:2064 weakly suppresses
  20. L12:965 weakly excites
  21. L20:2624 weakly excites
  22. L16:85 weakly excites
  23. L6:798 weakly suppresses
  24. L17:1416 weakly excites

Feeds later layers

  1. L27:612 weakly excites
  2. L27:1527 weakly excites
  3. L27:1560 weakly suppresses
  4. L26:1865 weakly excites
  5. L27:1282 weakly excites
  6. L27:1018 weakly excites unclear: mixed word fragments, no shared theme
  7. L26:2466 weakly excites
  8. L27:1242 weakly excites
  9. L26:2779 weakly excites
  10. L27:1298 weakly excites
  11. L27:1584 weakly suppresses
  12. L26:1158 weakly excites
  13. L26:2384 weakly excites
  14. L27:3067 weakly excites
  15. L26:831 weakly excites
  16. L27:983 weakly excites
  17. L26:2845 weakly excites
  18. L27:1252 weakly excites
  19. L27:1870 weakly excites
  20. L26:960 weakly suppresses
  21. L26:550 weakly suppresses
  22. L27:1845 weakly excites
  23. L27:1591 weakly excites
  24. L26:2667 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 -20.13 and -19.011 values between -14.53 and -13.421 values between -13.42 and -12.301 values between -12.30 and -11.181 values between -11.18 and -10.063 values between -10.06 and -8.943 values between -8.94 and -7.825 values between -7.82 and -6.718 values between -6.71 and -5.5913 values between -5.59 and -4.4727 values between -4.47 and -3.3539 values between -3.35 and -2.23138 values between -2.23 and -1.111510 values between -1.11 and 0.011920 values between 0.01 and 1.12118 values between 1.12 and 2.2436 values between 2.24 and 3.3624 values between 3.36 and 4.4816 values between 4.48 and 5.607 values between 5.60 and 6.728 values between 6.72 and 7.839 values between 7.83 and 8.954 values between 8.95 and 10.073 values between 10.07 and 11.192 values between 11.19 and 12.314 values between 12.31 and 13.431 values between 13.43 and 14.542 values between 14.54 and 15.661 values between 16.78 and 17.901 values between 20.14 and 21.251 values between 23.49 and 24.61offnegativepositiveoften

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. illery
  2. 观察
  3. hon
  4. щи
  5. clist
  6. wick
  7. ejs
  8. chos
  9. ôn
  10. obs

Suppresses

  1. izing
  2. aggio
  3. aggi
  4. aturing
  5. ization
  6. izational
  7. izado
  8. isation
  9. displayName
  10. 링