Neuron Explorer

Qwen3-0.6B

'to' after 'happen' (happen to)

A neuron in layer 5 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 word 'to' when it follows 'happen' or 'happened', as in 'if you happen to come upon'. Turning it up did not bring back that phrase, and the output wandered into unrelated topics.

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

    osity function to map wavelengths to the human eye's particular sensitivities. However, in the case of illuminance, there's only actual perception if your eye happens to be the surface in question.⏎I can basically grasp this, but then I come

  2. #2 peak

    .codeguru.com/algorithms/factorial.shtml I don't happen to think that the code there is easily understandable though...⏎January 10th,

  3. #3 peak

    umas, panthers, or cougars. Another name for them is catamount. These creatures are shy and isolated so they are seldom seen. If you happen to come upon one of these cougars, think smart. Cluster together with your hiking

  4. #4 peak

    s with lots of clear nasal drainage, but it can also cause laryngitis and bronchitis, and can trigger or worsen attacks if you happen to have asthma or other chronic lung problems. Since you do not stay completely immune to

  5. #5 peak

    -role fighter under any circumstance sounds like a bad idea, given that its primary function is air dominance, but every titan finds its challenger eventually. This one just happened to appear in the form of a bee swarm.⏎The drama unfolded Thursday at the Joint

  6. #6 peak

    firms can track the development of open source software through tools created for that purpose. There is a multitude that is available on the market, many of which also happen to be open source.⏎The Asian open source software movement has been gaining steam almost at

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: the user is a web application, and the browser is a problem. They are not isolated, and a new problem is happening. I am a web browser, and I need to be aware of something. The user has been working with a

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

    , cell and medical imaging are at a point comparable to the space<|endoftext|>SAFETY & ESCAPE PLAN⏎BEFORE A VIOLENT INCIDENT OCCURS⏎• Teach your children to call 911.⏎• Identify neighbors you

  2. #2 peak

    and that major hormonal changes are taking place in the body. The altered cycles are because of early menopause. However,

  3. #3 peak

    995 that gave Sweden permission to serve as America’s “protective power” in North Korea.⏎At the time, there were a lot of talks going on between U.S. and North Korean officials. The office was meant to serve as

  4. #4 peak

    án lab and first author of the article, explains that “for the first time we have a genetic model that allows us to understand the events that take place, starting from when cells begin to accumulate genomic errors until the development of a tum

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. L4:1292 suppresses
  2. L3:2540 excites
  3. L3:2683 suppresses
  4. L4:2748 weakly suppresses
  5. L4:1099 weakly suppresses
  6. L0:754 weakly suppresses
  7. L2:2984 weakly suppresses
  8. L3:2250 weakly suppresses
  9. L4:25 weakly excites
  10. L2:2516 weakly suppresses
  11. L2:942 weakly excites
  12. L3:2009 weakly suppresses
  13. L4:2415 weakly excites
  14. L2:1288 weakly excites
  15. L4:2150 weakly suppresses
  16. L4:2697 weakly excites
  17. L0:3007 weakly excites
  18. L4:1114 weakly suppresses
  19. L2:2398 weakly suppresses
  20. L4:2210 weakly suppresses
  21. L3:343 weakly suppresses
  22. L4:64 weakly suppresses
  23. L4:1936 weakly excites
  24. L4:37 weakly suppresses

Feeds later layers

  1. L6:2766 suppresses
  2. L10:1324 suppresses
  3. L11:2065 suppresses
  4. L10:1441 suppresses
  5. L9:1128 weakly excites
  6. L9:2428 weakly suppresses
  7. L22:1658 weakly excites
  8. L10:3050 weakly suppresses
  9. L20:1422 weakly excites
  10. L10:769 weakly suppresses
  11. L10:1838 weakly excites
  12. L25:2475 weakly excites
  13. L11:1723 weakly suppresses
  14. L10:204 weakly suppresses
  15. L15:1996 weakly excites
  16. L11:976 weakly excites
  17. L12:1609 weakly suppresses
  18. L7:2220 weakly suppresses
  19. L8:1689 weakly excites
  20. L10:698 weakly suppresses
  21. L7:2106 weakly excites
  22. L22:1290 weakly excites
  23. L16:2287 weakly suppresses
  24. L19:930 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 -2.57 and -2.481 values between -1.40 and -1.311 values between -1.31 and -1.221 values between -1.22 and -1.132 values between -1.13 and -1.042 values between -1.04 and -0.953 values between -0.95 and -0.863 values between -0.68 and -0.593 values between -0.59 and -0.509 values between -0.50 and -0.4116 values between -0.41 and -0.3238 values between -0.32 and -0.23186 values between -0.23 and -0.15530 values between -0.15 and -0.061438 values between -0.06 and 0.041008 values between 0.04 and 0.13439 values between 0.13 and 0.22151 values between 0.22 and 0.3050 values between 0.30 and 0.3917 values between 0.39 and 0.484 values between 0.48 and 0.572 values between 0.57 and 0.662 values between 0.66 and 0.751 values between 0.93 and 1.02offnegativepositiveoften

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. ader
  2. ˗
  3. 别人的
  4. Yük
  5. porter
  6. etheless
  7. ỏi
  8. олод
  9. oped
  10. ubern

Suppresses

  1. 发生
  2. iani
  3. _OCCURRED
  4. 万元以上
  5. Tac
  6. iage
  7. ilio
  8. 發生
  9. _occ
  10. commerce