Neuron Explorer

Qwen3-0.6B

'and' joining paired names or fields

A neuron in layer 25 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 lights up on the word 'and' when it links two named things, like two universities, two fields of study, or two researchers. When it was turned up, the model kept repeating the word 'beyond', a word close to 'other'. When it was turned down, the text became fill-in-the-blank style.

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

    of the nocturnal flyer, Ctenochasma elegans, a pterosaur, is shown.⏎Schmitz, a researcher in the Department of Evolution and Ecology at UC Davis, and colleague Ryosuke Motani made the determination by

  2. #2 peak

    the borders between anglophone and francophone countries.⏎- Colonial rule⏎ASJC Scopus subject areas⏎- Economics and Econometrics⏎- Organizational Behavior and Human Resource Management<|endoftext|>- About ASTI⏎- Pay and Conditions⏎- AST

  3. #3 peak

    better at deciphering.⏎Join us for an evening with two renowned researchers—Richard Alley, Evan Pugh Professor of Geosciences and an Associate of the Earth and Environmental Systems Institute at Penn State, and Gavin Schmidt, Director of the NASA God

  4. #4 peak

    combine facts, analysis, perspective, new ideas, and enthusiasm to make interesting and challenging topics highly readable.<|endoftext|>The MPI Multidimensional Poverty Index of the Oxford and the UN put the Ethiopia the 2 poorest country in the world while the Economist

  5. #5 peak

    Jacobean Drama to each Unit are provided in the list of UG-CBCS Central University & State University Syllabus eg. Dibrugarh University and Guwahati University so that you can easily browse through different College and University

  6. #6 peak

    contain that have been useful in research.⏎The glands are enclosed in a capsule of and internally divided into . Blood vessels and nerves enter the glands at the and gradually branch

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: [/u/], [/u/], [/u/], [/u/], [/u/], [/u/], [/u/], [/u/], [/u/], [/u

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

    More Info... by email⏎Consumer ethnocentrism is a psychological concept that refers to individuals who believe that their country's products are superior to those of other countries. This concept also describes consumers in one country thinking that purchasing products in

  2. #2 peak

    there is none worthy of worship except Allah. As we have previously mentioned, Allah is the personal name for the One God. In Arabic, Allah has no other meaning but One God! Anyone who believes

  3. #3 peak

    ) have led to this growth in the minority ethnic population, so that by 2011 around one in five people living in England identified with an ethnic group other than the majority White British. This growth

  4. #4 peak

    ⏎The finding, they say, also highlights one aspect of human cell biology that has not been replicated in other animal model systems. And is virtually impossible to use anything other than human embryonic stem cells to gather information of this kind.⏎One reason

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. L19:1754 excites
  2. L13:131 excites 'and' joining parallel phrases
  3. L20:952 suppresses
  4. L4:57 suppresses
  5. L9:92 suppresses "and" or "or" joining two items
  6. L24:2564 suppresses
  7. L19:2162 excites
  8. L21:1871 suppresses
  9. L24:492 suppresses
  10. L21:784 suppresses
  11. L16:376 suppresses 'or' joining alternatives
  12. L20:885 suppresses 'and' joining paired words
  13. L15:727 excites
  14. L18:1850 excites
  15. L19:2606 excites
  16. L11:406 suppresses
  17. L23:371 excites
  18. L17:1510 excites
  19. L19:2122 suppresses
  20. L10:74 excites
  21. L8:70 excites 'and' between stacked superlatives
  22. L17:572 excites
  23. L21:2296 excites
  24. L18:868 suppresses 'and' joining two nouns in a pair

Feeds later layers

  1. L27:922 excites
  2. L27:1608 weakly excites
  3. L26:124 weakly suppresses
  4. L27:1097 weakly excites
  5. L26:1728 weakly excites
  6. L27:2471 weakly excites
  7. L27:948 weakly excites
  8. L26:535 weakly excites
  9. L27:2723 weakly excites
  10. L27:712 weakly excites
  11. L27:2479 weakly excites
  12. L26:777 weakly excites
  13. L26:235 weakly suppresses
  14. L26:998 weakly suppresses
  15. L26:1615 weakly excites
  16. L26:1421 weakly suppresses
  17. L26:2627 weakly excites
  18. L26:499 weakly suppresses
  19. L27:1113 weakly excites
  20. L26:221 weakly suppresses
  21. L26:2073 weakly excites
  22. L27:1229 weakly excites
  23. L27:615 weakly suppresses
  24. L27:639 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 -15.22 and -14.242 values between -9.34 and -8.362 values between -8.36 and -7.383 values between -7.38 and -6.401 values between -6.40 and -5.412 values between -5.41 and -4.4313 values between -4.43 and -3.4514 values between -3.45 and -2.4757 values between -2.47 and -1.49382 values between -1.49 and -0.512783 values between -0.51 and 0.47417 values between 0.47 and 1.4577 values between 1.45 and 2.4334 values between 2.43 and 3.4131 values between 3.41 and 4.3915 values between 4.39 and 5.377 values between 5.37 and 6.3512 values between 6.35 and 7.3311 values between 7.33 and 8.316 values between 8.31 and 9.295 values between 9.29 and 10.274 values between 10.27 and 11.252 values between 11.25 and 12.237 values between 12.23 and 13.224 values between 13.22 and 14.202 values between 14.20 and 15.185 values between 15.18 and 16.161 values between 16.16 and 17.141 values between 17.14 and 18.123 values between 18.12 and 19.101 values between 19.10 and 20.081 values between 20.08 and 21.061 values between 22.04 and 23.021 values between 23.02 and 24.00offnegativepositiveoften

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. beyond
  2. 其他的
  3. 其他
  4. other
  5. other
  6. others
  7. 其他人
  8. Beyond
  9. related
  10. related

Suppresses

  1. ulative
  2. ี้
  3. quina
  4. saja
  5. �名
  6. imator
  7. gomery
  8. ść
  9. лист
  10. areth