Neuron Explorer

Qwen3-0.6B

unclear: opposite pairs and citation digits

A neuron in layer 20 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

This neuron is hard to pin down. It fires in places like 'alive (on) or dead (off)', 'from Z to A', and 'in like a lion, out like a lamb', where a word sets up its opposite. It goes strongly negative on digits inside citation page numbers. Turning it up gave muddled text about the sun.

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

    *) memchr (str, 'p', strlen(str));⏎printf ("'p' found at position %d.\

  2. #2 peak

    way's Game of Life⏎Conway's Game of Life consists of a world of cells on a grid. These cells are either alive (on) or dead (off). The state of cell is determined by a set of rules, which all

  3. #3 peak

    will make your list alphabetical A to Z. (Descending will reverse the list from Z to A).⏎There is an option of telling the program that your list has a header

  4. #4 peak

    that the same character is differently pronounced in different languages or even within the same, for example in English the character “u” is pronounced in “but” buht, yet also as in “

  5. #5 peak

    6th-18th, 2017, Vicenza, Italy)<|endoftext|>- Welcome to Botswana⏎- Where To Go⏎- What To Do⏎- Where To Stay⏎- Flora and Fauna⏎- Visitor

  6. #6 peak

    animal form she pleases; she first appears in the story as a dark brown moth.⏎How to pronounce æ?⏎The 'ae' pair, or the single 'æ' symbol mixed together, is not pronounced as two separate vowels. It comes

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 would like to provide a more comprehensive and ________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

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

    and Freedom”, Nature, Nov. 15, 1941, Vol. 148, pp. 580-584.⏎The reasons why the adoption

  2. #2 peak

    iofacial) area, and/or the genitals. "LEOPARD," an acronym for such characteristic abnormalities, stands for (L)entigines; (E)lectrocard

  3. #3 peak

    Increased with Fewer Antibiotics⏎Am Fam Physician. 2006 May 1;73(9):1619-1634.⏎Clinical Question: Does more cautious prescribing of antibiotics for sore throat increase the

  4. #4 peak

    ate the molecular basis of the genetic susceptibility to alcoholism.(JAMA. 1992;268:1877-1882)<|endoftext|>The Imaginative Treatment of Geometry in Waldorf Education⏎A

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. L19:2663 strongly suppresses
  2. L16:2181 strongly suppresses
  3. L17:1622 strongly excites
  4. L18:1781 strongly suppresses
  5. L14:1810 excites
  6. L19:2022 suppresses
  7. L13:2047 excites
  8. L17:2573 excites
  9. L17:1423 suppresses
  10. L19:1781 suppresses
  11. L18:1373 excites
  12. L19:2045 excites abbreviation pieces in organisation names
  13. L19:1829 suppresses opening bracket before an acronym
  14. L11:1153 suppresses
  15. L14:2419 excites
  16. L17:1590 suppresses
  17. L19:771 excites
  18. L17:2755 suppresses
  19. L18:1996 excites
  20. L17:1404 suppresses
  21. L12:1089 excites
  22. L18:772 excites
  23. L19:2772 excites
  24. L19:619 excites

Feeds later layers

  1. L22:530 strongly excites
  2. L27:129 strongly excites
  3. L23:238 strongly suppresses
  4. L25:652 excites
  5. L27:134 excites
  6. L21:1934 excites
  7. L21:111 excites
  8. L22:2241 suppresses
  9. L23:957 excites
  10. L21:144 suppresses
  11. L24:2215 excites
  12. L24:2767 excites
  13. L22:623 excites
  14. L27:583 excites
  15. L23:505 excites
  16. L26:1539 excites
  17. L26:788 suppresses
  18. L21:2623 excites
  19. L23:672 excites
  20. L27:1615 suppresses
  21. L27:1487 suppresses
  22. L22:839 suppresses
  23. L26:386 excites
  24. L24:142 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 -3.97 and -3.661 values between -3.66 and -3.351 values between -2.73 and -2.421 values between -2.42 and -2.112 values between -2.11 and -1.816 values between -1.81 and -1.5048 values between -1.50 and -1.19341 values between -1.19 and -0.88949 values between -0.88 and -0.571489 values between -0.57 and -0.26819 values between -0.26 and 0.05152 values between 0.05 and 0.3628 values between 0.36 and 0.6622 values between 0.66 and 0.9712 values between 0.97 and 1.289 values between 1.28 and 1.598 values between 1.59 and 1.902 values between 1.90 and 2.211 values between 2.21 and 2.522 values between 2.52 and 2.831 values between 2.83 and 3.131 values between 3.13 and 3.444 values between 3.44 and 3.752 values between 4.37 and 4.681 values between 5.30 and 5.601 values between 5.60 and 5.912 values between 6.22 and 6.531 values between 6.53 and 6.841 values between 8.07 and 8.38offnegativepositiveoften

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. 贷
  2. lock
  3. 岗
  4. 债
  5. 秤
  6. 瘤
  7. 관
  8. 税
  9. 梦
  10. مة

Suppresses

  1. 地理位置
  2. .googleapis
  3. 湖泊
  4. Nich
  5. 基金份额
  6. (directory
  7. 鸬
  8. 白银
  9. ParameterValue
  10. (filePath