Neuron Explorer

Qwen3-0.6B

'of' after a group noun

A neuron in layer 16 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 the word "of" when it follows a word for a group, like "groups of", "teams of", "packs of" or "sets of", where a list of members is about to come. When it was turned up, the model's writing drifted toward vague, wordy talk about people and groups, though the effect was not very clear.

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

    Teachers of Mathematics.⏎merical values and use it when x=3 (with success rates of 69% and 81% for the two groups of students, respectively). They were less successful, however, when asked to derive from

  2. #2 peak

    ative algebra, with the language and the problems of geometry.⏎The main objects of study in algebraic geometry are algebraic varieties, which are geometric manifestations of sets of solutions of systems of polynomial equations. Examples of the most studied classes of algebraic

  3. #3 peak

    classes serve as a legal cornerstone that prohibits discrimination in various aspects of life, particularly within the workplace.⏎Protected classes, in a legal context, refer to groups or individuals who are safeguarded by specific anti-discrimination laws, ensuring they are

  4. #4 peak

    fellows and who get on in life, and the men who recruit the aristocracy and who doubtless largely form the plutocracy. It is significant that the group of low-class men—artisans and peasants—and the men of religion, whose mission

  5. #5 peak

    usually smaller. They are generally about 5-7 feet long from nose to tail (the tail is usually about 2 feet long) and move in packs of up to 15 members! To learn more about gray wolves, you can

  6. #6 peak

    blindness. Two teams of scientists have used modified viruses injected into the eye to partially restore vision to people suffering

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 working on the first two of the three different ways I have seen the first two of the three different ways I have seen the first two of the three different ways... the first two of the three different ways I have seen the

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 podiatrist, and local podiatric societies and insurance companies may have directories of practitioners. Podiatrists are also listed in the yellow pages of your phone book. The State Board for Podiatry cannot refer you to a practitioner.⏎If

  2. #2 peak

    answer this question⏎Join a community of thousands of dedicated teachers and students.Join eNotes<|endoftext|>Is there something else you would like to know?⏎Use the alphabetical glossary listing to find a short definition to a medical or dental term. If you would

  3. #3 peak

    Named GLIMPSE360 (Galactic Legacy Mid-Plane Survey Extraordinaire project), the deep infrared survey captures only about 3% of the sky, but because it focuses on the plane of the Milky Way, where stars are

  4. #4 peak

    structive Sleep Apnea is a serious and potentially life-threatening sleep disorder in which blockage of the airway during sleep occurs. The muscles in the upper airway collapse, which interferes with proper breathing. The most noticeable

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. L10:1069 strongly excites
  2. L14:1823 suppresses
  3. L10:1318 excites
  4. L9:747 excites
  5. L11:1941 suppresses
  6. L6:511 excites
  7. L12:1191 excites
  8. L9:310 excites
  9. L12:1227 excites
  10. L4:2159 suppresses
  11. L12:1941 suppresses
  12. L5:1471 excites
  13. L13:2788 suppresses
  14. L14:1782 excites
  15. L12:957 excites
  16. L9:910 suppresses
  17. L13:2496 excites
  18. L10:1100 suppresses
  19. L12:1241 suppresses
  20. L9:504 suppresses
  21. L8:532 excites
  22. L15:2615 excites
  23. L10:1841 excites
  24. L4:920 excites

Feeds later layers

  1. L18:1098 excites
  2. L17:1246 excites
  3. L17:1157 excites
  4. L17:1390 suppresses
  5. L20:965 excites
  6. L18:968 excites
  7. L22:874 excites
  8. L18:919 excites
  9. L19:2473 excites
  10. L17:1535 suppresses
  11. L24:1547 weakly excites
  12. L27:2478 weakly excites
  13. L18:2429 weakly excites
  14. L17:2139 weakly excites
  15. L21:2373 weakly excites
  16. L24:361 weakly suppresses
  17. L23:1437 weakly excites
  18. L18:456 weakly suppresses
  19. L18:2621 weakly excites
  20. L18:1654 weakly suppresses
  21. L17:1274 weakly excites
  22. L26:1549 weakly excites
  23. L23:572 weakly excites
  24. L17:1873 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 -2.15 and -1.911 values between -1.41 and -1.1711 values between -1.17 and -0.92174 values between -0.92 and -0.681070 values between -0.68 and -0.431668 values between -0.43 and -0.19578 values between -0.19 and 0.06118 values between 0.06 and 0.3181 values between 0.31 and 0.5549 values between 0.55 and 0.8037 values between 0.80 and 1.0430 values between 1.04 and 1.2913 values between 1.29 and 1.5319 values between 1.53 and 1.787 values between 1.78 and 2.028 values between 2.02 and 2.279 values between 2.27 and 2.529 values between 2.52 and 2.762 values between 2.76 and 3.015 values between 3.01 and 3.255 values between 3.25 and 3.502 values between 3.50 and 3.741 values between 3.74 and 3.992 values between 3.99 and 4.231 values between 4.23 and 4.481 values between 4.73 and 4.972 values between 5.95 and 6.202 values between 6.69 and 6.942 values between 7.43 and 7.67offnegativepositiveoften

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. 里的
  2. 成员
  3. 里
  4. containing
  5. members
  6. members
  7. 裡
  8. 中新
  9. 中的
  10. Members

Suppresses

  1. اته
  2. 程
  3. iento
  4. _unicode
  5. _CLICKED
  6. rito
  7. PropertyChanged
  8. еле
  9. reads
  10. 旒