Neuron Explorer

Qwen3-0.6B

"of" after member, version, or part nouns

A neuron in layer 7 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 lights up on the word "of" when it follows a noun like "members", "versions", "ends" or "aspects", as in "members of the family". It switches off on "in" after years or events, such as "founding, in 1817". Turning it up made the model's writing wander into unrelated topics, so the steering did not show a clear theme.

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

    like a miniature version of its full-sized species.⏎And to achieve this look, you’ll need to maintain

  2. #2 peak

    it? What is the setting? What happens? What do you like/not like about the way the author writes?⏎Word by word⏎Similar but more formal versions of this should be happening at school, but parents can reinforce this learning at home.

  3. #3 peak

    an impartial humanitarian organization. It uses the information obtained to identify civilians and members of a paramilitary force belonging to a specific ethnic group in order to arrest or kill

  4. #4 peak

    environment. From their lineage to their habitat, spiny king crabs are extremely interesting crustaceans.⏎Spiny king crabs are invertebrates and members of the king

  5. #5 peak

    all people, and many members of our society are still striving to achieve them.⏎“We hope this Convention will be followed

  6. #6 peak

    (Buckminster-Fuller, 1969)⏎The buckyball is the simplest of the carbon structures known as fullerenes. Members of the fullerene family are a major subject of nanotechnology research. 3-D

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: of the year, 2001-2004-05-06-07-88-88-87-77-71-75-77-73-

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

    colony on African soil. The success of Cuffee’s venture led to the founding, in 1817, of the American Colonization Society, with the mission

  2. #2 peak

    8 filing of a Citizen Submission with the Commission for Environmental Cooperation, an institution formed to fulfill the environmental provisions of NAFTA. The Citizen Submission culminated in the publication in 2002 of a report documenting the risk to human health that tox

  3. #3 peak

    ”, a translation of Fr. Boccalini’s “Del Ragguagli di Parnasso”, 1612. Beginning with the fourth edition in 1615, the third Rosicrucian rudiment, “

  4. #4 peak

    they return to Hiva with the news that indeed,<|endoftext|>The intent of this work is five fold: 1) To spatially characterize and monitor the distribution, abundance, and size of

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. L5:246 weakly suppresses
  2. L2:2451 weakly excites
  3. L5:1721 weakly excites
  4. L0:2630 weakly excites
  5. L3:155 weakly suppresses
  6. L3:1107 weakly suppresses
  7. L4:2853 weakly excites
  8. L6:1735 weakly suppresses
  9. L0:2028 weakly excites
  10. L2:2458 weakly excites
  11. L2:2772 weakly excites
  12. L3:1548 weakly suppresses
  13. L5:2115 weakly excites
  14. L2:2670 weakly suppresses
  15. L5:21 weakly excites
  16. L2:946 weakly suppresses
  17. L4:23 weakly excites
  18. L4:2448 weakly suppresses numerical computing and library names
  19. L0:695 weakly excites
  20. L2:428 weakly suppresses
  21. L0:554 weakly excites
  22. L2:3028 weakly excites
  23. L5:239 weakly suppresses
  24. L1:2065 weakly excites

Feeds later layers

  1. L8:202 suppresses
  2. L9:1210 suppresses
  3. L8:1350 excites
  4. L22:142 excites 'of' after a quantity or noun phrase
  5. L18:953 suppresses
  6. L16:961 suppresses
  7. L15:2736 suppresses
  8. L8:526 suppresses
  9. L13:1942 suppresses
  10. L8:1867 suppresses immediately repeated words
  11. L17:1771 suppresses
  12. L27:658 suppresses
  13. L26:711 suppresses 'years' after an age number
  14. L23:567 weakly suppresses
  15. L15:1474 weakly suppresses
  16. L12:2641 weakly suppresses
  17. L20:2599 weakly suppresses
  18. L19:1080 weakly suppresses
  19. L25:2020 weakly suppresses
  20. L25:1497 weakly excites
  21. L22:288 weakly suppresses
  22. L15:894 weakly suppresses
  23. L24:2219 weakly excites
  24. L15:2143 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 -1.42 and -1.292 values between -0.92 and -0.793 values between -0.79 and -0.676 values between -0.67 and -0.5421 values between -0.54 and -0.4198 values between -0.41 and -0.29344 values between -0.29 and -0.161005 values between -0.16 and -0.041716 values between -0.04 and 0.09509 values between 0.09 and 0.22134 values between 0.22 and 0.3427 values between 0.34 and 0.4711 values between 0.47 and 0.597 values between 0.59 and 0.724 values between 0.72 and 0.844 values between 0.84 and 0.973 values between 0.97 and 1.103 values between 1.10 and 1.223 values between 1.35 and 1.472 values between 1.47 and 1.602 values between 1.60 and 1.721 values between 1.72 and 1.851 values between 2.73 and 2.861 values between 3.48 and 3.61offnegativepositiveoften

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. 记者在
  2. 人死亡
  3. 这些问题
  4. 市政府
  5. 省政府
  6. 人在
  7. ьев
  8. DeltaTime
  9. elerik
  10. 他是一个

Suppresses

  1. of
  2. _of
  3. -of
  4. Of
  5. Of
  6. OF
  7. of
  8. oks
  9. _OF
  10. của