Neuron Explorer

Qwen3-0.6B

comma before a trailing phrase

A neuron in layer 21 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 comma that ends a full statement and is followed by a phrase adding more detail, like ", blazing a trail" or ", allowing developers". It tends to be negative on similar commas in technical process descriptions. Turning it up gave text that did not clearly carry this pattern forward.

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

    aschke⏎University News Service⏎September 26, 2011⏎Victor Frankenstein’s infamous monster led a brief, tragic existence, blazing a trail of death and destruction that prompted mobs of angry villagers to take up

  2. #2 peak

    (Minnesota) for three years before her marriage to Dr. Will Kennicott, of Gopher Prairie. She is a rebel against ugliness and conformity, and one factor in her decision to accept Kennicott is the opportunity to make

  3. #3 peak

    Roman Catholic faith and speaking a medieval Hungarian dialect, they still live in relative isolation, harvesting vegetables and nuts, some of which they exchange for oil, rice and other

  4. #4 peak

    ello at the Covent Garden theatre. The prospects of fame, fortune, and freedom were very appealing to young Booth, and he displayed remarkable talent from an early age, deciding on a career in the theatre by the age of 17. He

  5. #5 peak

    as developers navigate through various programming languages and frameworks to create powerful and intuitive applications.⏎The Significance of Coding in App Development⏎Coding is the backbone of app development, allowing developers

  6. #6 peak

    in mythology, the stars⏎Orion is one of the easier to recognizable and noticeable constellations. It is perhaps the most splendid of all constellations, befitting for a venerated character from ancient times. Named for a hunter

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: For the first of two years, I have for the first of two years for the first of two of them for the first of two of them for the first of two of then. For the first of two of then for the first of

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

    amount of Ar-Cl2 or N2-Cl2, pure Ar, N2 gas into the bottom of the aluminum melt through the high-speed rotating rotor, and the impeller is fully chopped to form dispersed bubbles to make the aluminum liquid

  2. #2 peak

    oth surface, based on the generation principle of spherical involute tooth surface, the reverse motion of the generated motion of the tooth surface is taken as the cutting motion, and the reduced spiral tooth surface with spherical involute tooth shape can be milled

  3. #3 peak

    rotary movement to the linear movement as far as possible. Based on the analysis of the tooth surface cutting movement, the revolution and rotation of the blade are further analyzed, as shown in the figure.<|endoftext|>Source: The Independent – independent.co.uk⏎

  4. #4 peak

    all about making a 'tray' and covering it with paper.⏎My cousin was 24 years old on Friday April 5th in 1974. On what day⏎of the week was she born?⏎A tetrom

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. L15:184 excites
  2. L9:207 suppresses
  3. L9:46 suppresses
  4. L18:233 suppresses
  5. L18:1103 excites
  6. L11:840 suppresses
  7. L10:370 weakly suppresses
  8. L10:155 weakly excites
  9. L20:2998 weakly excites
  10. L10:873 weakly excites
  11. L20:510 weakly excites
  12. L9:281 weakly excites
  13. L11:1602 weakly suppresses
  14. L20:476 weakly excites commas that lead into a continuation
  15. L7:92 weakly excites
  16. L20:1466 weakly suppresses
  17. L8:498 weakly suppresses
  18. L19:2058 weakly suppresses
  19. L11:223 weakly excites
  20. L2:763 weakly excites
  21. L18:768 weakly excites
  22. L18:843 weakly suppresses
  23. L7:90 weakly suppresses
  24. L10:662 weakly excites

Feeds later layers

  1. L24:140 weakly excites
  2. L22:1425 weakly excites
  3. L22:609 weakly excites
  4. L23:424 weakly excites
  5. L27:1134 weakly suppresses
  6. L27:1620 weakly excites
  7. L24:1443 weakly suppresses
  8. L23:1639 weakly excites
  9. L27:1628 weakly suppresses
  10. L25:591 weakly excites
  11. L23:772 weakly excites
  12. L26:252 weakly excites
  13. L27:410 weakly excites
  14. L22:506 weakly suppresses
  15. L24:2880 weakly suppresses
  16. L25:376 weakly excites
  17. L23:2082 weakly excites
  18. L23:1155 weakly suppresses
  19. L25:1425 weakly suppresses
  20. L24:2425 weakly suppresses
  21. L22:2204 weakly excites
  22. L26:733 weakly excites
  23. L25:744 weakly suppresses
  24. L23:2249 weakly 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 -5.18 and -4.601 values between -4.60 and -4.023 values between -2.86 and -2.293 values between -2.29 and -1.7115 values between -1.71 and -1.13118 values between -1.13 and -0.551997 values between -0.55 and 0.031620 values between 0.03 and 0.6197 values between 0.61 and 1.1912 values between 1.19 and 1.7612 values between 1.76 and 2.345 values between 2.34 and 2.925 values between 2.92 and 3.502 values between 3.50 and 4.082 values between 4.08 and 4.652 values between 4.65 and 5.235 values between 5.23 and 5.811 values between 5.81 and 6.393 values between 6.39 and 6.971 values between 8.12 and 8.701 values between 10.44 and 11.011 values between 13.33 and 13.911 values between 17.38 and 17.95offnegativepositiveoften

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. being
  2. being
  3. erville
  4. тер
  5. uffling
  6. enance
  7. averse
  8. czy
  9. ère
  10. /Main

Suppresses

  1. лим
  2. 那就是
  3. IFICATIONS
  4. aturas
  5. 那么
  6. |#
  7. arden
  8. atura
  9. иф
  10. 那就