Neuron Explorer

Qwen3-0.6B

unclear: word fragments and digits inside numbers

A neuron in layer 18 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

The snippets don't share a clear pattern. The strongest hits sit on odd fragments inside words or numbers, such as the piece "Sevent" in "Seventeen" or "them" in "thematically". It is strongly negative on the digit "7" inside years like 1765. Turning it up or down gave unrelated text, so the neuron's job is hard to pin down.

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

    highlighted for years as the ninth planet in the solar system.⏎However, in 20

  2. #2 peak

    000 and were so interconnected by marriage that they represented a giant family rather than a new polyglot community. The community was governed by The Council of Seventeen in Amsterdam, who governed the far-reaching empire of the Dutch East India Company

  3. #3 peak

    shape them using a jigsaw. 3. Drill holes in the stringers and attach

  4. #4 peak

    cholesterol from animal products we take in when eating. Even more interesting is the fact that at

  5. #5 peak

    changing fundamentally, but also the way people learn. Learning trends such as mobile learning, micro

  6. #6 peak

    heat. Fish oils are notorious for their self-heating properties. Mineral oils like white spirit, mineral turpentine or lubricating oil are not prone to self-heating and will not cause spontaneous ignition.⏎It

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 time when a different team is the first to the final, but the game is already been decided, and the game is already 78% of the way, and time is 3:35, but the game is

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

    th century, the term was extended from the talker to the conversation of such persons. Th<|endoftext|>Radicalism and the Common People: Britain c.1789-1880 - 30 credits (HS17

  2. #2 peak

    7, 28, 31, 32, 33, 34, 35, 36, 37, 39, 49, 51, 6

  3. #3 peak

    became legal tender in much of the country and many used hemp to pay their taxes.⏎The cultivation and spread of hemp in the United States continued and in 1765 George Washington made reference to growing hemp in

  4. #4 peak

    is predicated on changing human behavior; the purview of Psychologists (reviewed in , see also [2,3,4,5,6,7]).⏎So-called “environmental problems” are really problems of human behavior, caused

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. L2:2028 strongly suppresses
  2. L0:2319 strongly excites
  3. L4:1049 strongly excites
  4. L17:350 excites
  5. L10:1210 suppresses
  6. L14:643 excites
  7. L0:756 suppresses
  8. L16:17 suppresses
  9. L6:1434 suppresses
  10. L1:19 suppresses
  11. L16:673 excites
  12. L17:389 suppresses
  13. L16:112 excites
  14. L16:503 excites
  15. L17:127 suppresses the word 'would' before a verb
  16. L7:922 suppresses
  17. L16:113 suppresses
  18. L15:80 suppresses
  19. L16:120 suppresses
  20. L11:981 excites
  21. L16:56 suppresses
  22. L15:82 suppresses
  23. L17:391 suppresses
  24. L16:84 excites

Feeds later layers

  1. L19:1269 strongly suppresses the word 'about'
  2. L20:1793 suppresses
  3. L20:558 suppresses
  4. L22:607 excites
  5. L19:551 excites
  6. L20:342 excites
  7. L20:121 excites
  8. L19:683 excites
  9. L19:316 excites
  10. L21:2906 excites
  11. L20:2115 excites
  12. L19:2293 excites
  13. L22:2379 excites
  14. L23:2739 suppresses
  15. L26:2860 excites
  16. L26:1063 excites
  17. L19:576 excites
  18. L22:961 excites
  19. L22:2526 excites
  20. L20:1228 excites
  21. L19:1615 excites
  22. L20:1310 excites
  23. L20:183 excites
  24. L20:2371 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 -28.16 and -27.401 values between -23.63 and -22.871 values between -22.87 and -22.121 values between -19.85 and -19.101 values between -16.08 and -15.321 values between -12.30 and -11.551 values between -10.79 and -10.041 values between -7.77 and -7.023 values between -6.27 and -5.512 values between -4.76 and -4.002 values between -4.00 and -3.252 values between -3.25 and -2.493 values between -2.49 and -1.7410 values between -1.74 and -0.9836 values between -0.98 and -0.233590 values between -0.23 and 0.53223 values between 0.53 and 1.2829 values between 1.28 and 2.04offnegativepositiveoften

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. nad
  2. Ỗ
  3. adium
  4. utilisateur
  5. éal
  6. INGS
  7. عار
  8. →elem
  9. iores
  10. enburg

Suppresses

  1. 庸
  2. 放心
  3. truncate
  4. TripAdvisor
  5. <hr
  6. ⏎ ⏎
  7. 脸
  8. reel
  9. 厘
  10. ⏎ ⏎