Neuron Explorer

Qwen3-0.6B

"of" or "for" after minister or ministry titles

A neuron in layer 8 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 small words "of" or "for" that come right after a government job title, like "Minister of Health" or "Minister for Housing". Turned up, the model's output became rambling and repetitive, not clearly about government.

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

    a a third time, this time in a shorter version with only <|endoftext|>ATHENS – The digital exhibition Proving History, presented by the Hellenic Ministry of

  2. #2 peak

    vice-chancellor, probably overwhelmed by the occasion, asked the minister of state for petroleum and natural gas and the minister of state for external affairs to help the university procure an Army tank for the campus. Clearly,

  3. #3 peak

    70 Elementary Schooling Act was the efficient begin of state-financed training in England and Wales. The Secretary of State initially simply took the place of the Minister for Science as chair of the various Privy Council committees for civil science. It emphasizes

  4. #4 peak

    .⏎problem6. Compare and contrast a peer-to-peer network architecture to Client-Server architecture. Support your answer with proper diagrams and exs.<|endoftext|>The Minister for Housing, Local Government and Heritage, Darragh O’Brien, and the Minister

  5. #5 peak

    emy laws (Salman Taseer, the former governor of Punjab, and Shahbaz Bhatti, the Federal Minister for Minorities) have been assassinated. Since 1990,

  6. #6 peak

    rights and freedoms,” National Congress of Australia’s First Peoples co-chair Rod Little said.⏎Assistant Minister for Social Services and Multicultural Affairs Minister Zed Seselja described the UN criticisms

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: The question is, what is the answer to this problem? If $ x $ is 2, then $ x - 1 = 2 - 1 = 2 $, then $ x - 3 = 2 $

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

    Central Committee of the Bolshevik party.⏎The Bolshevik party were a minority but gained publicity after being the only

  2. #2 peak

    Science practices meet their potential in bridging the world’s STI gaps and enabling sustainable development. The Member States of the Organisation have examined this issue in the UNESCO Executive Board:⏎- Preliminary study of the technical, financial and legal aspects on the

  3. #3 peak

    of a species of fish (Mitsukurina owstoni) from Japan, the type of a distinct family of lamnoid sharks". Proceedings of the California Academy of Sciences (Series 3) Zoology 1 (6):

  4. #4 peak

    researchers have demonstrated that it is not. In 1970, dentist Wallace Johnson and others published the results of their research in Journal of the American Dental Association. The report indicated that "mercury, an essential element 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. L7:1131 excites
  2. L3:1237 excites
  3. L6:1742 weakly excites
  4. L3:1336 weakly suppresses
  5. L4:581 weakly excites
  6. L7:2067 weakly excites
  7. L6:2775 weakly excites
  8. L7:1780 weakly excites
  9. L3:1919 weakly suppresses timestamps after dates
  10. L6:2579 weakly excites
  11. L7:1049 weakly suppresses
  12. L4:2674 weakly excites
  13. L3:1851 weakly suppresses
  14. L7:2994 weakly excites
  15. L4:2063 weakly suppresses
  16. L5:1156 weakly excites
  17. L0:901 weakly suppresses 'this' after sentence-initial 'In'
  18. L7:1083 weakly excites
  19. L4:2365 weakly excites
  20. L4:2229 weakly suppresses
  21. L4:2617 weakly suppresses
  22. L5:1146 weakly excites
  23. L3:2671 weakly suppresses
  24. L2:3066 weakly suppresses

Feeds later layers

  1. L9:1505 excites
  2. L12:1612 excites
  3. L16:1855 excites
  4. L13:1181 excites 'of' after a definition or noun
  5. L16:2715 excites
  6. L13:1150 excites
  7. L15:2810 excites
  8. L20:697 excites
  9. L11:1246 excites
  10. L13:1651 excites
  11. L18:2179 excites
  12. L11:2298 excites
  13. L12:2737 excites
  14. L13:2665 excites
  15. L9:1601 suppresses
  16. L13:408 excites
  17. L20:1850 weakly excites
  18. L23:1321 weakly excites
  19. L19:2827 weakly excites
  20. L16:1380 weakly excites
  21. L17:2907 weakly excites
  22. L9:663 weakly suppresses
  23. L16:2216 weakly excites
  24. L9:2190 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.

5 values between -1.35 and -0.81243 values between -0.81 and -0.263582 values between -0.26 and 0.2862 values between 0.28 and 0.826 values between 0.82 and 1.372 values between 1.37 and 1.913 values between 1.91 and 2.463 values between 2.46 and 3.001 values between 3.00 and 3.541 values between 19.85 and 20.39offnegativepositiveoften

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. ervice
  2. serviço
  3. overseeing
  4. orea
  5. Svc
  6. cope
  7. .getComponent
  8. kitchens
  9. .getService
  10. Servi

Suppresses

  1. изм
  2. �新
  3. flyer
  4. quotes
  5. cff
  6. figcaption
  7. ---</
  8. ,filename
  9. 份
  10. 램