Neuron Explorer

Qwen3-0.6B

dash starting a new list item

A neuron in layer 8 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 a dash that starts a new bullet or point in a list of advice or facts, after a finished sentence. It also responds weakly to words like 'third' or 'Second' that begin a count in a list. Turning it up gave off-topic text, so the steering output does not confirm the pattern.

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

    cholesterol and diabetes. Exercise also reduces the risk of stroke, promotes a sense of well-being, and reverses the ill effects of chronic stress on the body.⏎- Being intellectually and socially engaged – There appears to be some real benefit from engaging in

  2. #2 peak

    this year. Getting in the habit of meeting the criteria that is expected beyond the formal education years is an ideal way to make sure you are working toward success.⏎-

  3. #3 peak

    careful however, because if the two grandparents had the same first name, sometimes two children who were named after them ended up with the same first and last name!⏎- If all the grandparents, previous spouses and deceased children were

  4. #4 peak

    , biographies,⏎etc.). Ask your librarian or counselor about award-winning books that are recommended⏎for readers your age and include at least one of those titles.⏎- Find the books

  5. #5 peak

    for game-changers amongst those who wear multiple hats, and should model game changing for our students through experiential opportunities that do not have proscribed outcomes.⏎- They are

  6. #6 peak

    replacement level (which is currently approximately 2.1 children per woman). This means that on average, couples are no longer having enough children to replace them.⏎- This low-fertility regime resulted from the fall in the total fertility rate that followed

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 been thinking about lately, and I have been thinking about lately. I am sleeping. I am sleeping. I am sleeping. I am sleeping. I am sleeping. I am sleeping. I am sleeping. I am sleeping. I

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

    majority of artists living there. First, she was a single young female in a society that basically frowned upon such status. Second, she was financially independent, and third, she had received a typical "proper" education as behooved a young

  2. #2 peak

    of a royal widow, Jacquetta Duchess of Bedford, and a minor gentleman, Sir Richard Woodville. Elizabeth’s first husband, was Sir John Grey, and they had two sons before he died at the Second Battle of St Albans.

  3. #3 peak

    now one of the most common urban birds in its range, we provide the first assessment of two contrasting urban colonization models. First, that urbanization occurred independently. Second, that following initial urbanization, urban-adapted individuals colonized other urban

  4. #4 peak

    the Woman's Board of Missions, located in Boston (WBM), the Woman's Board of Missions of the Interior, located in Chicago (WBMI), the Woman's Board of Missions for the Pacific, located in San Francisco (

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. L7:837 suppresses
  2. L6:2534 suppresses
  3. L6:74 suppresses
  4. L7:333 excites
  5. L4:91 excites
  6. L6:478 excites
  7. L5:965 suppresses
  8. L6:393 suppresses
  9. L7:1223 excites "and" joining two decades or times
  10. L6:2012 suppresses
  11. L3:865 excites
  12. L6:137 suppresses
  13. L5:2400 suppresses
  14. L5:313 excites
  15. L6:2522 excites
  16. L6:1005 suppresses
  17. L7:610 excites
  18. L7:337 excites
  19. L5:696 excites
  20. L6:1630 weakly excites
  21. L5:305 weakly suppresses
  22. L3:47 weakly suppresses
  23. L6:286 weakly excites
  24. L6:79 weakly excites

Feeds later layers

  1. L14:718 excites
  2. L9:404 excites
  3. L10:295 excites
  4. L9:187 excites
  5. L16:443 excites
  6. L9:733 excites
  7. L9:322 suppresses
  8. L17:884 excites
  9. L12:228 suppresses end-of-document token after web boilerplate
  10. L10:605 excites
  11. L18:1219 excites
  12. L9:113 suppresses
  13. L16:612 suppresses
  14. L9:889 excites
  15. L10:1185 excites
  16. L19:821 excites
  17. L10:490 excites
  18. L15:981 excites
  19. L16:578 excites
  20. L10:703 suppresses
  21. L9:44 excites
  22. L9:342 excites
  23. L18:937 excites
  24. L9:251 suppresses

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 -2.48 and -2.381 values between -2.28 and -2.192 values between -1.80 and -1.701 values between -1.70 and -1.612 values between -1.61 and -1.514 values between -1.41 and -1.313 values between -1.22 and -1.124 values between -1.12 and -1.026 values between -1.02 and -0.936 values between -0.93 and -0.8313 values between -0.83 and -0.7316 values between -0.73 and -0.6414 values between -0.64 and -0.5427 values between -0.54 and -0.4448 values between -0.44 and -0.3464 values between -0.34 and -0.25178 values between -0.25 and -0.15454 values between -0.15 and -0.051348 values between -0.05 and 0.041012 values between 0.04 and 0.14462 values between 0.14 and 0.24174 values between 0.24 and 0.3446 values between 0.34 and 0.4314 values between 0.43 and 0.532 values between 0.53 and 0.631 values between 0.63 and 0.721 values between 0.72 and 0.821 values between 0.92 and 1.011 values between 1.11 and 1.212 values between 1.31 and 1.40offnegativepositiveoften

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. NSS
  2. orama
  3. utions
  4. another
  5. another
  6. anders
  7. 另一个
  8. _maps
  9. устройств
  10. Programs

Suppresses

  1. .listen
  2. ิก
  3. Cc
  4. 状
  5. 郎
  6. ชะ
  7. しかない
  8. .rb
  9. 日夜
  10. 名叫