Neuron Explorer

Qwen3-0.6B

end-of-document token after web boilerplate

A neuron in layer 12 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

The neuron lights up on the invisible 'end of text' marker that appears where one web page or article stops and another starts, especially after citation lines and site footers. Turning it up gave off-topic text and did not show a clear meaning.

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

    , words or ideas that would upset a reasonable person in any culture.⏎This resource is licensed by LoubeeSav⏎under the iSLCollective Copyright License<|endoftext|>|Does a 'proper' tragic hero need a crown?|⏎Christmas Eve brought

  2. #2 peak

    The above story is based on materials provided by NASA's Marshall Space Flight Center. Note: Materials may be edited for content and length.⏎Cite This Page:<|endoftext|>By A. M. Sullivan⏎CHAPTER IX. (continued)⏎From the Atlas

  3. #3 peak

    He points out the effort of educational institutes who have created awareness and encouraged the students to report<|endoftext|>Farmers in designated Nitrate Vulnerable Zones (NVZs) are to have the amount of manure they can spread on fields slashed as Def

  4. #4 peak

    issues."⏎Materials provided by Indiana University-Purdue University Indianapolis School of Science. Note: Content may be edited for style and length.⏎Cite This Page:<|endoftext|>Briers, Mississippi facts for kids⏎Quick facts for kids⏎|Elevation

  5. #5 peak

    this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with focus by next letter<|endoftext|>I was listening to Brooke Gladstone's ON THE MEDIA podcast about the difficulty of

  6. #6 peak

    110,000 wildlife and science products⏎Multi-currency. Secure worldwide shipping⏎Wildlife, science and conservation since 1985<|endoftext|>A new study recently published showed that for women who are overweight, moderate weight loss

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 can't be sure. But I can't get to the top. That's not the problem, but I don't know, I can't get to the top. Wait, that's not the problem, maybe it's something 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

    a thousand years to decompose.⏎Scientist strengthened a cow milk protein, casein, with clay and glyceraldehyde (a triose monosaccharide). They chose casein, already popular for use in adhesives, because

  2. #2 peak

    Palliser, dans la Terre de Kerguelen [on sheet with] Plan du Havre de Noel, dans la Terre de Kerguelen, by Cook/Benard, from Atlas du Troisi<|endoftext|>One of

  3. #3 peak

    puritan society. In his novel,The Scarlet Letter, Nath<|endoftext|>What they learned after only seven years is worth remembering (or learning, as the case may be).⏎The United States has operated under two constitutions. The first, The Articles

  4. #4 peak

    ,000 die.⏎- MRSA: A Superbug Super Villain This is MRSA, or methicillin-resistant Staphylococcus aureus, a pervasive superbug that causes infections ranging from mild skin infections to serious infections

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. L10:1177 strongly excites
  2. L9:192 strongly suppresses
  3. L6:145 strongly excites
  4. L9:136 strongly excites
  5. L10:228 strongly suppresses
  6. L4:317 strongly excites
  7. L4:69 strongly excites
  8. L10:295 strongly suppresses
  9. L9:700 suppresses
  10. L11:112 suppresses
  11. L9:94 suppresses
  12. L1:459 excites
  13. L8:334 suppresses
  14. L4:535 suppresses
  15. L9:322 suppresses
  16. L6:62 suppresses
  17. L9:235 suppresses
  18. L9:236 suppresses
  19. L3:721 suppresses
  20. L8:454 suppresses dash starting a new list item
  21. L10:443 excites
  22. L9:72 suppresses
  23. L2:222 excites
  24. L9:10 suppresses

Feeds later layers

  1. L18:142 suppresses
  2. L14:718 excites
  3. L15:306 excites
  4. L17:884 excites
  5. L19:821 excites
  6. L16:578 excites
  7. L26:823 excites
  8. L16:443 excites
  9. L16:612 weakly suppresses
  10. L15:981 weakly excites
  11. L14:931 weakly suppresses
  12. L14:1998 weakly excites
  13. L14:2895 weakly excites
  14. L18:625 weakly excites
  15. L14:553 weakly excites
  16. L16:413 weakly suppresses
  17. L14:933 weakly excites
  18. L16:2994 weakly excites
  19. L15:2277 weakly excites
  20. L25:1149 weakly excites
  21. L18:1219 weakly excites
  22. L16:417 weakly suppresses
  23. L14:137 weakly suppresses
  24. L23:2859 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 -4.47 and -4.313 values between -3.32 and -3.161 values between -3.16 and -2.991 values between -2.99 and -2.832 values between -2.83 and -2.672 values between -2.67 and -2.501 values between -2.50 and -2.346 values between -2.34 and -2.173 values between -2.17 and -2.018 values between -2.01 and -1.848 values between -1.84 and -1.6810 values between -1.68 and -1.5219 values between -1.52 and -1.3513 values between -1.35 and -1.1929 values between -1.19 and -1.0248 values between -1.02 and -0.8651 values between -0.86 and -0.6982 values between -0.69 and -0.53124 values between -0.53 and -0.37192 values between -0.37 and -0.20323 values between -0.20 and -0.04703 values between -0.04 and 0.13760 values between 0.13 and 0.29611 values between 0.29 and 0.46417 values between 0.46 and 0.62234 values between 0.62 and 0.79124 values between 0.79 and 0.9571 values between 0.95 and 1.1135 values between 1.11 and 1.2816 values between 1.28 and 1.446 values between 1.44 and 1.612 values between 1.77 and 1.942 values between 1.94 and 2.10offnegativepositiveoften

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.  
  2.   
  3. ệu
  4. mond
  5. vant
  6. amel
  7.  
  8. doses
  9. avan
  10. phận

Suppresses

  1. ��
  2. 窠
  3. aracı
  4. onomy
  5. ering
  6. brig
  7. ὗ
  8. itus
  9. arty
  10. ожно