Neuron Explorer

Qwen3-0.6B

"and" or "or" joining two items

A neuron in layer 9 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 responds to the words "and" and "or" when they link two things in a pair, like "he and she" or "Mr. and Mrs.". When it was turned up, the model's sentences started repeating and leaning on "the weather was cold" type phrasing, which is not clearly about joining words.

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

    one be puzzled? Because there’s one he and there’s one she and life is pronominally simple.⏎But things get tricky when she and she or he and he or singular, non-binary they and they rent huts:⏎Sandra and

  2. #2 peak

    , T., Boesman, P., Olarte, L., Salaman, P. & Olaciregui , C.|⏎|Facilitator/Compiler(s):||Benstead, P., Capper, D., Isher

  3. #3 peak

    8-76-1Purchased with funds contributed by Mr. and Mrs. Isaac Tatnall Starr, 1928⏎LabelThis

  4. #4 peak

    , and other books of Jewish interest. Owing to his sympathy with the Reformers during

  5. #5 peak

    , and the test laboratories were built during the past five years.⏎Each laboratory is linked through

  6. #6 peak

    , and scan additional documents.⏎Should I transfer the rights?⏎When you create something unlikely to

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 a new project. I have a person. I have a thing. This is the process. The method. The system. The way. The way. The way. I have a situation. and a feeling. and a thought

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

    ](verb): to make bare; strip; denude.⏎This fun word comes from the latin word “denudatus” which means “naked” or “ bare.”⏎With the days getting colder, the deciduous

  2. #2 peak

    7,000 births and the risk increases with maternal age. The conditions are associated with severe mental handicap and several physical defects. Most affected individuals die before or soon after birth and they rarely survive beyond the first year of life.⏎Today we

  3. #3 peak

    large degree age-related, and may also be more developed in highly trained racehorses.⏎|Number of pages||7|⏎|Journal||Veterinarija ir Zootechnika|⏎|Publication status||Published - 17 Oct 2

  4. #4 peak

    ies, even if abandoned, in ruined condition or transformed for other use, also served as symbolic mezuzahs to mark towns

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. L4:57 strongly suppresses
  2. L5:880 strongly excites
  3. L8:70 excites 'and' between stacked superlatives
  4. L4:108 excites "or" offering a choice in web text
  5. L6:427 suppresses
  6. L3:1951 suppresses
  7. L0:313 excites "and" joining paired nouns
  8. L1:1250 suppresses
  9. L4:2363 excites
  10. L3:250 excites
  11. L5:548 excites
  12. L4:2270 excites
  13. L0:2146 suppresses
  14. L6:584 weakly suppresses
  15. L7:42 weakly excites
  16. L1:1710 weakly suppresses
  17. L2:2969 weakly suppresses
  18. L1:1597 weakly suppresses
  19. L8:1083 weakly excites "x" or "by" between two dimensions
  20. L4:1330 weakly suppresses
  21. L1:2120 weakly excites
  22. L6:1706 weakly suppresses
  23. L7:836 weakly excites
  24. L5:1469 weakly suppresses

Feeds later layers

  1. L17:572 strongly suppresses
  2. L13:131 strongly suppresses 'and' joining parallel phrases
  3. L18:870 strongly suppresses
  4. L16:376 strongly suppresses 'or' joining alternatives
  5. L15:727 strongly suppresses
  6. L10:74 strongly suppresses
  7. L12:51 excites
  8. L17:1537 suppresses
  9. L11:876 suppresses 'other' after 'among' in lists
  10. L17:2274 suppresses
  11. L15:2765 suppresses
  12. L16:2577 excites
  13. L17:2332 suppresses 'and' joining two clauses
  14. L17:3061 suppresses
  15. L14:1116 suppresses 'and' inside lists
  16. L16:2580 suppresses 'and' linking food or health nouns
  17. L15:1501 suppresses "and" joining two parts of a name or phrase
  18. L19:2080 suppresses ampersand joining two items
  19. L16:1669 suppresses
  20. L16:2442 suppresses
  21. L20:885 suppresses 'and' joining paired words
  22. L14:801 suppresses 'and' or 'or' joining list items
  23. L13:125 suppresses
  24. L25:1694 suppresses 'and' joining paired names or fields

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 -9.69 and -9.421 values between -9.42 and -9.161 values between -8.63 and -8.371 values between -8.11 and -7.841 values between -7.84 and -7.581 values between -7.58 and -7.321 values between -7.32 and -7.052 values between -7.05 and -6.792 values between -6.79 and -6.522 values between -6.52 and -6.264 values between -6.26 and -6.001 values between -6.00 and -5.732 values between -5.47 and -5.213 values between -5.21 and -4.943 values between -4.94 and -4.686 values between -4.68 and -4.415 values between -4.41 and -4.154 values between -4.15 and -3.895 values between -3.89 and -3.622 values between -3.62 and -3.363 values between -3.36 and -3.105 values between -3.10 and -2.839 values between -2.83 and -2.577 values between -2.57 and -2.312 values between -2.31 and -2.0410 values between -2.04 and -1.787 values between -1.78 and -1.517 values between -1.51 and -1.257 values between -1.25 and -0.998 values between -0.99 and -0.727 values between -0.72 and -0.4610 values between -0.46 and -0.20271 values between -0.20 and 0.072533 values between 0.07 and 0.33871 values between 0.33 and 0.60103 values between 0.60 and 0.86offnegativepositiveoften

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. 质量和
  2. 企业和
  3. 时间和
  4. 建设和
  5. elle
  6. 和个人
  7. 意大
  8. 经验和
  9. and
  10. 平民

Suppresses

  1. /or
  2. rog
  3. 其它
  4. nbsp
  5. 珅
  6. /OR
  7. ients
  8. дей
  9. ifice
  10. reas