Neuron Explorer

Qwen3-0.6B

commas inside lists of items

A neuron in layer 19 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 fires on a comma that separates items in a list, such as 'strengths, weaknesses' or 'science, technology'. Turning it up made the model produce broken, repetitive text, so the steering results say little about it.

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

    These include physical features of the<|endoftext|>Learn vocabulary, terms, and more with flashcards, games, and other study tools. Optional

  2. #2 peak

    developers.⏎Key Factors for Successful App Development⏎Several key factors contribute to successful app development through coding. First and foremost, a clear understanding of the app’s purpose, target audience, and desired

  3. #3 peak

    to each other.⏎- Identify points where team members make decisions⏎- Note whether all members of the team are involved.<|endoftext|>The acronym SWOT stands for strengths, weaknesses, opportunities and threats. Organizations of all sizes use SWOT analysis to assess

  4. #4 peak

    ry disease characteristics, and the clinical parameters.⏎The investigators scoured the Pubmed, EMBASE, and CINAHL databases from their inception to February of

  5. #5 peak

    of Scouts BSA.⏎Printable helps for requirements:⏎Get started on your Next Big Thing Nova award with these ideas and related achievements:⏎STEM stands for science, technology, engineering, and mathematics. The NOVA Awards program combines knowledge of STEM

  6. #6 peak

    adaptogenic remedy. Ashwagandha, an Indian<|endoftext|>Moral FriendSeries teasers are where you try to complete the sequence of a series of letters, numbers or objects.⏎Explain why these words were put in the following sequence:⏎

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 a 500000 in my now my now now now now now now in my now now now now now now now now now my now now now now now now now now now now my now now now now now

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

    xia nervosa will typically fall under one of two forms of self-starvation: restricting, in which the person participates in extreme diets, fasting, or exercise; or binge-eating to purging, in which the person eats a large amount of

  2. #2 peak

    one<|endoftext|>Basic Statistics Tools⏎In this workshop, you'll learn basic concepts from descriptive statistics including mean, median, mode, standard deviation and normal distribution as well as common statistical symbols. You'll learn

  3. #3 peak

    s from the company of people and worldy affairs, either permanently, as a hermit or monk; or partially, as every member of a religious institute; or temporarily,k as in a retreat. Physical solitude is not escapism or isolationism

  4. #4 peak

    Dr Nicole Highet says. “In addition, there are behavioural benefits to do with goal-setting and a sense of achievement in mastering a particular task. There are also social benefits

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. L14:830 suppresses
  2. L14:312 excites
  3. L7:8 excites
  4. L14:313 weakly suppresses
  5. L6:296 weakly suppresses
  6. L16:896 weakly excites
  7. L14:126 weakly suppresses
  8. L3:189 weakly suppresses
  9. L7:207 weakly excites
  10. L3:250 weakly excites
  11. L5:169 weakly suppresses
  12. L1:591 weakly suppresses
  13. L16:269 weakly suppresses
  14. L11:2015 weakly suppresses
  15. L10:554 weakly excites
  16. L4:108 weakly excites "or" offering a choice in web text
  17. L3:186 weakly excites
  18. L4:1227 weakly excites
  19. L6:429 weakly suppresses
  20. L16:2151 weakly excites
  21. L18:1154 weakly suppresses
  22. L16:1659 weakly excites
  23. L14:1664 weakly excites
  24. L15:6 weakly suppresses

Feeds later layers

  1. L20:1677 suppresses
  2. L20:955 weakly suppresses
  3. L20:1372 weakly suppresses
  4. L21:2355 weakly suppresses
  5. L20:385 weakly suppresses
  6. L20:2276 weakly suppresses
  7. L22:270 weakly excites
  8. L21:1886 weakly suppresses
  9. L20:1292 weakly excites
  10. L21:580 weakly suppresses
  11. L27:2288 weakly excites
  12. L25:541 weakly excites
  13. L20:2016 weakly excites
  14. L24:337 weakly suppresses
  15. L21:225 weakly excites
  16. L21:1837 weakly suppresses
  17. L26:1752 weakly excites
  18. L26:613 weakly excites
  19. L26:596 weakly excites
  20. L22:2417 weakly suppresses
  21. L27:668 weakly suppresses
  22. L20:659 weakly suppresses
  23. L20:194 weakly excites
  24. L27:3066 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 -11.70 and -11.181 values between -10.65 and -10.121 values between -10.12 and -9.593 values between -9.59 and -9.071 values between -9.07 and -8.543 values between -8.54 and -8.011 values between -8.01 and -7.484 values between -7.48 and -6.967 values between -6.96 and -6.439 values between -6.43 and -5.9019 values between -5.90 and -5.3715 values between -5.37 and -4.8524 values between -4.85 and -4.3234 values between -4.32 and -3.7955 values between -3.79 and -3.2677 values between -3.26 and -2.7482 values between -2.74 and -2.21150 values between -2.21 and -1.68226 values between -1.68 and -1.15349 values between -1.15 and -0.63764 values between -0.63 and -0.101124 values between -0.10 and 0.43378 values between 0.43 and 0.96213 values between 0.96 and 1.49159 values between 1.49 and 2.0187 values between 2.01 and 2.5457 values between 2.54 and 3.0729 values between 3.07 and 3.6020 values between 3.60 and 4.128 values between 4.12 and 4.652 values between 5.18 and 5.711 values between 5.71 and 6.232 values between 6.76 and 7.291 values between 8.34 and 8.871 values between 8.87 and 9.40offnegativepositiveoften

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. okies
  2. Submitted
  3. eil
  4. Mosque
  5. illes
  6. 族自治州
  7. ?>>
  8. 型企业
  9. 岁以上
  10. ?</

Suppresses

  1. finally
  2. .gnu
  3. wg
  4. PREC
  5. myself
  6. ierung
  7. dag
  8. definit
  9. cw
  10. spath