Neuron Explorer

Qwen3-0.6B

closing parentheses in lists

A neuron in layer 24 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 closing bracket that ends a short note in a list or table, like the ")" after a measurement, a year range, or a format name. Changing it did not produce a clear theme in the generated text.

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

    |Length:||205 ft (62.48 m)o/a|⏎|Beam:||33 ft (10.06 m)|⏎|Draught:||11.5 ft (3.5

  2. #2 peak

    , sugar|⏎|vak, vaacha||voice, vocal|⏎|veda, vidya (from vid) (knowledge)||wit, wise (from vid)|⏎The postfix ‘er’ in words indicates the occupation / action of a<|endoftext|>

  3. #3 peak

    |miti (measure)||metric, meter, measure (latin metiri)|⏎|gouh (earth)||geometry, geography (greek geos)|⏎|lok (place)||location, locus, locale|⏎|ek (one

  4. #4 peak

    ategorize this paper)|⏎|Buy the book||$8.32 used (63% off) $17.25 new (22% off) $21

  5. #5 peak

    ' Classroom.|⏎|Levels:||Late Elem. (3-5), Middle School (6-8), High School (9-12)|⏎|Resource Types:||Public and Government Institutions, Web Interactive/Java|⏎

  6. #6 peak

    nose, nostril, nasal|⏎|danta (tooth)||dental, denture|⏎|janu||knee (french genou)|⏎|pad (feet)||pedicure, pedestrian, centipede

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 goal is to make the world a home for all people.\ \ Now it seems that some people might justifiar'tc that the goal is a bad option because some people might not be able to self-resource themselves (they

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

    ) testifies (Adv. Har. xxix.1) It is derived from the Hebrew Ebion, "poor," and was not given, as Origen supposes, for their low view of Christ. Id. at 6

  2. #2 peak

    . However, right around 235, Hippolytus was reconciled with the Church.⏎One of the other items on Pontian’s plate was Origen. As a young man, Origen established a school of philosophy in Alexandria

  3. #3 peak

    37, has been selected as an Associate Laureate in the Rolex Awards for an ambitious project to encourage more than 500 families in the areas around Orurillo and Pucará to broaden the genetic variety of their crops. More

  4. #4 peak

    download these videos from Media Hopper Create.⏎[Featured image: US Flag Backlit, CC BY SA, Jnn13, Wikimedia Commons]<|endoftext|>Coronary artery

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. L19:118 strongly suppresses
  2. L0:304 strongly suppresses
  3. L16:39 strongly excites
  4. L16:97 excites
  5. L16:0 suppresses
  6. L4:489 suppresses
  7. L19:1636 excites closing bracket after a list number or letter
  8. L16:13 suppresses
  9. L22:866 suppresses
  10. L16:9 excites
  11. L16:61 excites
  12. L20:61 suppresses
  13. L18:569 excites
  14. L16:2 suppresses
  15. L2:1808 excites
  16. L19:233 excites
  17. L20:441 suppresses
  18. L2:1834 excites
  19. L23:1167 excites
  20. L21:173 excites
  21. L16:17 excites
  22. L16:94 suppresses
  23. L2:75 suppresses
  24. L4:412 suppresses

Feeds later layers

  1. L26:362 strongly excites
  2. L26:381 strongly excites
  3. L26:3057 excites
  4. L27:1305 suppresses
  5. L27:1687 suppresses
  6. L27:2174 suppresses
  7. L26:725 suppresses
  8. L27:2250 suppresses
  9. L26:236 weakly suppresses
  10. L27:429 weakly excites
  11. L25:1312 weakly suppresses unclear: mixed, Chinese and name fragments
  12. L27:613 weakly suppresses
  13. L25:483 weakly excites
  14. L26:1864 weakly suppresses
  15. L27:2714 weakly suppresses
  16. L25:422 weakly suppresses
  17. L27:657 weakly excites
  18. L27:1147 weakly suppresses
  19. L27:139 weakly excites
  20. L27:145 weakly suppresses dots inside web addresses
  21. L26:1138 weakly suppresses
  22. L25:244 weakly excites
  23. L26:193 weakly suppresses
  24. L27:1113 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 -4.12 and -3.371 values between -3.37 and -2.6210 values between -2.62 and -1.8763 values between -1.87 and -1.12369 values between -1.12 and -0.373439 values between -0.37 and 0.386 values between 0.38 and 1.133 values between 2.63 and 3.381 values between 5.63 and 6.381 values between 6.38 and 7.133 values between 8.63 and 9.382 values between 10.88 and 11.631 values between 11.63 and 12.381 values between 13.13 and 13.881 values between 13.88 and 14.631 values between 14.63 and 15.381 values between 17.63 and 18.381 values between 22.88 and 23.632 values between 23.63 and 24.381 values between 25.13 and 25.88offnegativepositiveoften

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. ribbon
  2. iscrim
  3. isce
  4. cluded
  5. ris
  6. clusion
  7. vol
  8. 看书
  9. onium
  10. atten

Suppresses

  1. ,
  2. ,在
  3. cq
  4. ,并
  5. ،
  6. ,《
  7. //
  8. cj
  9. tement
  10. statt