Neuron Explorer

Qwen3-0.6B

comma after an introductory phrase

A neuron in layer 18 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 wakes up on a comma that comes right after a short opening phrase, such as 'In some people,' or 'For many organizations,' where the main sentence is about to start. When it was turned up, the model's text drifted into unrelated, rambling sentences.

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

    And it will move you into alignment with Judaism’s most noble values.⏎The Origin of Epidemics⏎A little history is in order.⏎Throughout most of human history, there were no epidemic diseases.⏎“No one got the flu, not even the common

  2. #2 peak

    In some people, however, cells start to ignore insulin, which means the pancreas must manufacture more

  3. #3 peak

    1764, and he was the first person officially authorized to practice law in Georgia.⏎From 1777 until 1868, when the Constitution of 1868 made the name "attorney general

  4. #4 peak

    , Cultism, and Multiculturalism. Multicultural Education, 8(3), 2-14.<|endoftext|>Data is life⏎For many organizations, data is the lifeblood. It’s what keeps the business kicking and breathing.

  5. #5 peak

    2% to 6% of schoolchildren, mostly adolescent girls. It is most commonly diagnosed between ages 13 and 15.⏎In some children, symptoms begin after a triggering event, such as physical trauma, surgery, infection or

  6. #6 peak

    shaped using bronze tools, then smoothed with sand in order to fit tightly together. No mortar was used in the construction. Even after many centuries of wear and tear, the precision is amazing; It would be impossible to slide a piece of paper between

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 I, I am, I have, I have, I have, I have, I I I I I I I I I I I I I I I I I I I I I I I I I I I I 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

    prompt asks you for an extension to be appended to the rectified images.⏎The next prompt will ask you whether a camera angle map should be produced and if yes, what should be its name.⏎After that you are asked if overwriting existing maps

  2. #2 peak

    and specific mental ability tests are important tools for vocational guidance. Researchers are now asking whether performance on such tests is based on differences in brain structure, and if so, can brain scans

  3. #3 peak

    - but look at the C++-tagged posts in this blog for examples of what the compiler thinks isn't a bad idea.)⏎if (x=0) init_subsystem();⏎So lots of books have been written about how to write code

  4. #4 peak

    are affecting data quality? If so, what sources?⏎- What makes a duplicate a duplicate?⏎- Should records be merged

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. L17:1426 excites
  2. L15:280 weakly excites
  3. L9:1579 weakly suppresses
  4. L6:2843 weakly suppresses
  5. L0:561 weakly suppresses
  6. L16:1886 weakly excites
  7. L15:2038 weakly suppresses
  8. L5:30 weakly excites
  9. L16:1072 weakly excites
  10. L2:622 weakly excites
  11. L15:122 weakly suppresses
  12. L4:308 weakly excites
  13. L15:136 weakly suppresses
  14. L4:1981 weakly excites
  15. L15:78 weakly excites
  16. L6:687 weakly suppresses
  17. L13:3018 weakly suppresses
  18. L17:944 weakly excites
  19. L17:1039 weakly suppresses
  20. L10:2648 weakly excites
  21. L16:711 weakly suppresses
  22. L15:89 weakly excites
  23. L13:91 weakly excites
  24. L10:332 weakly suppresses

Feeds later layers

  1. L19:1979 suppresses
  2. L21:881 suppresses comma after 'if' clause in prose, braces in code
  3. L24:405 suppresses
  4. L27:1779 suppresses
  5. L21:1811 suppresses
  6. L24:880 suppresses
  7. L27:967 weakly suppresses
  8. L20:122 weakly excites
  9. L19:973 weakly excites
  10. L22:470 weakly suppresses
  11. L25:386 weakly suppresses
  12. L25:6 weakly suppresses
  13. L21:434 weakly suppresses
  14. L22:689 weakly suppresses comma closing an intro clause
  15. L19:1598 weakly excites comma before a qualifying clause
  16. L26:2964 weakly excites
  17. L23:292 weakly suppresses
  18. L19:2866 weakly excites
  19. L19:624 weakly excites
  20. L20:2430 weakly suppresses
  21. L19:2352 weakly suppresses
  22. L26:377 weakly suppresses 'sh' starting a word piece
  23. L20:206 weakly suppresses
  24. L20:2412 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 -5.80 and -5.421 values between -4.26 and -3.881 values between -1.95 and -1.577 values between -1.57 and -1.1810 values between -1.18 and -0.80214 values between -0.80 and -0.412149 values between -0.41 and -0.031381 values between -0.03 and 0.3693 values between 0.36 and 0.7412 values between 0.74 and 1.137 values between 1.13 and 1.513 values between 1.51 and 1.903 values between 1.90 and 2.285 values between 2.28 and 2.671 values between 2.67 and 3.052 values between 3.44 and 3.824 values between 4.21 and 4.592 values between 4.59 and 4.981 values between 4.98 and 5.361 values between 5.36 and 5.752 values between 5.75 and 6.133 values between 6.13 and 6.521 values between 7.67 and 8.052 values between 8.05 and 8.441 values between 8.82 and 9.211 values between 9.21 and 9.59offnegativepositiveoften

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. abbr
  2. 文化和
  3. .band
  4. UUID
  5. ания
  6. ание
  7. /response
  8. estead
  9. ertext
  10. arged

Suppresses

  1. {⏎
  2. {⏎⏎
  3. {⏎
  4. {⏎⏎
  5. { ⏎ ⏎
  6. {⏎⏎⏎
  7. ({⏎
  8. { ⏎ ⏎
  9. { ⏎
  10. { ⏎