Neuron Explorer

Qwen3-0.6B

ellipsis dots after a cut-off quote

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 fires on the first dot of an ellipsis (". . .") where a quote or sentence trails off mid-thought. When it was turned up, the model wrote long strings of dots.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    over time. In the early church, believers focused upon genealogies so much that Paul had to warn his readers twice (!) not to “give heed to .

  2. #2 peak

    up the surface with thin layers of pigment, the color of the underlying layers modifying those on top. Painter Gino Severini reported that "Matisse said . . . that everything that did not contribute to the balance and rhythm of [this

  3. #3 peak

    15, 2003, the National⏎Transportation Safety Board (NTSB) made its boldest statement to date.⏎They stated that “ . . . the safe operation of fifteen-passenger⏎vans requires a knowledge and

  4. #4 peak

    and local adoption agencies follow here at home?⏎The present system of screening and approving textbooks dates to pre-war Japan: "Struggles over the national narrative existed . . . before and even during World War II, when official narratives such as the

  5. #5 peak

    ms we read, “Give thanks to the Lord, for he is good; his love endures forever. Let the redeemed of the Lord tell their story. . . . Let them give thanks to

  6. #6 peak

    lizards!⏎Ecologist Sean Doody discovered this while in the field with student Phillip Paull. The two were lifting eggs out of a crevice when . . . Holy Dooley! The hatchlings literally popped out and sprinted away!⏎

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: how many times a couple wants to have a child 10 years away? How many times a couple wants to have a child 20 years away? How many times a couple wants to have a child 50 years away?

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

    ’t require any qualifications and is open to anyone.⏎The course includes videos of farmer John Geldard from Plumgarth’s Farm Shop and Café in Kendal, members of

  2. #2 peak

    .⏎Metsäpohjainen piha paransi päiväkotilasten immuunijärjestelm

  3. #3 peak

    . Indirect reciprocity is a social interaction in which one actor helps another and is then benefited by a third party. Gossip has also been identified by Robin Dunbar, an evolutionary biologist, as aiding social bonding in large groups.⏎The word

  4. #4 peak

    built environment, she is the author of Bulldozer: Demolition and Clearance of the Postwar Landscape, winner of the 2017 Lewis Mumford Prize for the best book in American planning history. She is particularly interested in

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. L23:641 strongly suppresses
  2. L22:748 strongly suppresses 'and so on' list endings
  3. L22:701 strongly excites
  4. L22:2065 excites
  5. L18:2722 suppresses
  6. L18:2185 suppresses
  7. L17:2866 suppresses
  8. L19:1265 suppresses
  9. L17:2209 excites
  10. L18:2938 excites
  11. L16:3039 suppresses
  12. L21:2493 excites
  13. L23:1171 weakly suppresses
  14. L15:3052 weakly suppresses
  15. L18:742 weakly excites
  16. L19:831 weakly suppresses
  17. L16:76 weakly suppresses
  18. L4:1468 weakly suppresses
  19. L7:412 weakly excites
  20. L9:1533 weakly suppresses
  21. L18:1118 weakly excites
  22. L13:611 weakly excites
  23. L24:2698 weakly suppresses
  24. L21:475 weakly excites

Feeds later layers

  1. L27:1182 strongly excites
  2. L26:1825 excites the first dot of an ellipsis
  3. L27:409 suppresses
  4. L27:703 excites
  5. L26:481 weakly suppresses
  6. L27:2054 weakly suppresses
  7. L26:626 weakly excites
  8. L26:943 weakly excites
  9. L26:1453 weakly suppresses
  10. L26:887 weakly excites
  11. L27:2562 weakly suppresses
  12. L27:2222 weakly excites
  13. L26:2399 weakly excites
  14. L26:174 weakly excites
  15. L27:977 weakly suppresses
  16. L26:2489 weakly suppresses
  17. L26:2547 weakly suppresses
  18. L26:948 weakly excites
  19. L27:2989 weakly suppresses
  20. L26:2182 weakly excites
  21. L27:2343 weakly excites capital N starting a name or word
  22. L26:593 weakly excites
  23. L26:167 weakly excites
  24. L27:1627 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 -14.82 and -14.051 values between -14.05 and -13.291 values between -11.76 and -10.991 values between -5.62 and -4.862 values between -4.86 and -4.094 values between -4.09 and -3.324 values between -3.32 and -2.564 values between -2.56 and -1.7997 values between -1.79 and -1.031410 values between -1.03 and -0.262169 values between -0.26 and 0.51101 values between 0.51 and 1.2744 values between 1.27 and 2.0423 values between 2.04 and 2.8111 values between 2.81 and 3.5710 values between 3.57 and 4.345 values between 4.34 and 5.113 values between 5.11 and 5.874 values between 5.87 and 6.643 values between 6.64 and 7.413 values between 7.41 and 8.171 values between 8.17 and 8.942 values between 8.94 and 9.711 values between 9.71 and 10.472 values between 12.77 and 13.541 values between 15.07 and 15.84offnegativepositiveoften

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. .
  2. /.
  3. (.
  4. `.
  5. ."
  6. .'
  7. ,.
  8. (.
  9. ._
  10. [.

Suppresses

  1. cd
  2. (cd
  3. ld
  4. cd
  5. →cd
  6. cds
  7. CD
  8. _cd
  9. CD
  10. CDs