Neuron Explorer

Qwen3-0.6B

"and" in subject-area lists

A neuron in layer 14 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 "and" inside short lists of topics or keywords, such as "Physics and Astronomy" in a list of subject areas or "Ceramics and Composites". Turning it up gave text that did not show this list pattern, so the steering adds little.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).

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

    |Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most computer users are familiar with the concept of spy

  2. #2 peak

    experimentally derived films. Good agreement between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Ceramics and Composites⏎- Condensed Matter Physics⏎- Materials Chemistry<|endoftext|>Humans are intimately

  3. #3 peak

    original location and eventually came to the Museum.⏎Social Tags [?]european art 1100-1500 [x] european decorative arts and sculpture [x] fransiscan monestary too [x] of

  4. #4 peak

    in a discontinuity analysis on observations near 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

  5. #5 peak

    African American material and in particular the Civil Rights struggle.⏎In the History: UK section you can find out about the British Empire (Captives: Britain, Empire and The World 1600-1850, Linda Colley

  6. #6 peak

    he was appointed the head of KHAD, the secret police. Under Najibullah's control, it is claimed that KHAD arrested, tortured and executed tens of thousands of Afghans. Najibullah replaced Babrak

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 can't be in a place like a place where a person who who... something... something... I... something... something... I can't be... I can't be... I can't be... I can't be... something...

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

    http://www.cal-ipc.org/ip/management/plant_profiles/Rumex_acetosella.php⏎Rumex acetosella (red sorrel, sheep sorrel)⏎Rumex acetosella (red sorrel, sheep

  2. #2 peak

    , physical destruction (curettage with eletrodesiccation) and Mohs Micrographic Surgery. Newer methods under investigation include photodynamic therapy and immunochemotherapy.⏎In 1935, Dr. Frederic

  3. #3 peak

    11). http://www.cdc<|endoftext|>|This article is about the assassin of pharaoh Cleopatra. For other uses, see Aya or Amun.|⏎Prior to establishing the Hidden Ones, Aya served as an

  4. #4 peak

    Similar to assets, liabilities can also be classified as current or long-term, depending on their repayment timeline.⏎Shareholders' equity, also referred to as net worth or owner's equity, represents the residual interest in the company's assets after deducting

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. L12:2981 suppresses
  2. L12:1756 excites
  3. L10:74 excites
  4. L9:92 suppresses "and" or "or" joining two items
  5. L8:70 excites 'and' between stacked superlatives
  6. L4:57 suppresses
  7. L13:131 excites 'and' joining parallel phrases
  8. L10:1643 suppresses
  9. L12:1235 suppresses commas inside numbers and lists
  10. L13:1982 suppresses
  11. L11:406 suppresses
  12. L12:2071 excites
  13. L5:1469 suppresses
  14. L12:51 excites
  15. L13:371 suppresses
  16. L6:427 suppresses
  17. L13:977 suppresses
  18. L6:1630 excites
  19. L12:2057 suppresses
  20. L7:999 weakly suppresses
  21. L12:2409 weakly excites
  22. L9:738 weakly suppresses
  23. L12:1095 weakly suppresses
  24. L12:3006 weakly suppresses

Feeds later layers

  1. L16:1774 suppresses
  2. L15:631 suppresses
  3. L15:1340 suppresses
  4. L16:2733 suppresses
  5. L15:217 excites
  6. L17:84 excites
  7. L16:300 suppresses commas inside lists of nouns
  8. L16:2900 suppresses
  9. L15:2287 suppresses
  10. L15:1636 excites
  11. L17:2002 suppresses
  12. L15:2400 suppresses
  13. L21:735 suppresses 'or' in ranges and alternatives
  14. L15:2534 excites
  15. L15:1064 weakly suppresses
  16. L15:55 weakly suppresses
  17. L16:857 weakly excites
  18. L16:1346 weakly suppresses
  19. L17:3061 weakly suppresses
  20. L16:2209 weakly excites
  21. L16:717 weakly suppresses
  22. L16:1234 weakly suppresses
  23. L17:1234 weakly suppresses
  24. L15:124 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 -5.12 and -4.881 values between -3.66 and -3.421 values between -2.93 and -2.691 values between -2.69 and -2.443 values between -2.44 and -2.202 values between -2.20 and -1.961 values between -1.96 and -1.714 values between -1.71 and -1.476 values between -1.47 and -1.226 values between -1.22 and -0.988 values between -0.98 and -0.7421 values between -0.74 and -0.49149 values between -0.49 and -0.251801 values between -0.25 and -0.011733 values between -0.01 and 0.24115 values between 0.24 and 0.4822 values between 0.48 and 0.7212 values between 0.72 and 0.975 values between 0.97 and 1.218 values between 1.21 and 1.463 values between 1.46 and 1.701 values between 1.70 and 1.942 values between 1.94 and 2.191 values between 2.19 and 2.431 values between 4.38 and 4.62offnegativepositiveoften

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. ựa
  2. ution
  3. olars
  4. 的学习
  5. S
  6. concentration
  7. 的责任
  8. ut
  9. 善
  10. ется

Suppresses

  1. azo
  2. agrams
  3. ombat
  4. robe
  5. obook
  6. convertible
  7. otron
  8. ву
  9. ditor
  10. theon