Neuron Explorer

Qwen3-0.6B

the word 'list' and list-like nouns

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 responds to the word 'list' and to nouns that name a list or a set of items, such as 'species list' or 'the following list'. Turning it up did not give a clear, list-related change in the output.

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

    's steamship. […]⏎2008 Americas Most Endangered Historic Places Two locations in New York State have been listed on the National Trust for Historic Preservation's annual list of America's Most Endangered Places. The non-profit membership organization […]

  2. #2 peak

    : hills, woods⏎and lakes. Furthermore they boast very important works of art like frescos and statues.⏎UNESCO World Heritage Committee has put the Site on its List for the following reasons:⏎The realization of an architectural and sacred art’s work

  3. #3 peak

    an essay.⏎3. Make a list of the characteristics of the educated person as presented in the Commitment. Identify where

  4. #4 peak

    so on, in industries related to food, drugs, military equipment, manufacturing, computer equipment, software, household appliances, floor coverings, and paint, just to name a few. However, one area that's historically

  5. #5 peak

    Inc.⏎Some researchers have found that the following factors increase your chance of developing HG. If you have any of these risk factors, tell your doctor:⏎The following list of symptoms are general and may be caused by other, less serious health conditions.

  6. #6 peak

    easy as with other programming languages.⏎To create a list of integers, create a new variable and:⏎That’s simple right? And that

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 concept of a branchingancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheancheanche

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

    with the arrival of the new national director of TNC, Julio Cárcamo. However the current status of the project is unclear.⏎Main policy lessons⏎Tulio Monterroso, AMUPROLOAGO: mailto:firstname.lastname

  2. #2 peak

    f/f; corresponds to the Dirac delta function, δ(t), the unit element of the set of generalized functions.⏎(To be continued.)⏎HOME PAGE OF Thayer Watkins<|endoftext|>Pakistan is a Republic founded in 1947 and located

  3. #3 peak

    youth, citizens and policy makers are coming together to find solutions for creating a sustainable society.⏎Take a sustainability course this semester!⏎Looking to incorporate sustainability into your courseload? For historical comparison, UAF offered 157 sustainability focused and

  4. #4 peak

    published online today in Open Biology⏎, the open-access journal of Great Britain’s Royal Society.This image compares glioblastoma cells untreated or treated with antisense oligonucleotides (ASO) that modulate splicing for PK-M

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. L21:2773 strongly suppresses
  2. L24:594 suppresses
  3. L20:686 excites
  4. L23:710 excites
  5. L17:917 excites
  6. L22:1151 excites
  7. L22:740 excites
  8. L17:1947 suppresses
  9. L17:1944 suppresses
  10. L18:2067 excites
  11. L20:1756 excites
  12. L18:2465 excites
  13. L18:968 excites
  14. L23:1103 suppresses
  15. L17:3028 excites
  16. L17:1789 suppresses
  17. L18:1256 suppresses
  18. L13:2788 suppresses
  19. L17:1436 suppresses
  20. L19:1909 weakly excites
  21. L17:2555 weakly excites
  22. L17:1727 weakly excites
  23. L22:1369 weakly excites
  24. L23:3013 weakly excites

Feeds later layers

  1. L27:989 strongly suppresses
  2. L27:1462 suppresses
  3. L27:1804 weakly suppresses
  4. L26:2931 weakly suppresses
  5. L26:2383 weakly suppresses
  6. L27:2202 weakly suppresses
  7. L27:2138 weakly suppresses
  8. L27:1122 weakly suppresses
  9. L27:1688 weakly suppresses
  10. L27:1344 weakly suppresses
  11. L26:829 weakly excites
  12. L26:1381 weakly excites
  13. L26:112 weakly excites
  14. L27:2684 weakly suppresses
  15. L27:2577 weakly suppresses
  16. L26:805 weakly excites
  17. L27:1395 weakly suppresses
  18. L27:1581 weakly excites
  19. L26:2258 weakly suppresses
  20. L27:2114 weakly excites
  21. L27:824 weakly suppresses
  22. L26:2683 weakly suppresses
  23. L26:1779 weakly excites
  24. L26:1077 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.

2 values between -8.78 and -7.431 values between -6.08 and -4.742 values between -4.74 and -3.3910 values between -3.39 and -2.0473 values between -2.04 and -0.693630 values between -0.69 and 0.66160 values between 0.66 and 2.0118 values between 2.01 and 3.362 values between 3.36 and 4.701 values between 4.70 and 6.054 values between 7.40 and 8.752 values between 8.75 and 10.101 values between 12.79 and 14.141 values between 26.28 and 27.631 values between 43.81 and 45.16offnegativepositiveoften

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. uate
  2. از
  3. ️
  4. ес
  5. ogui
  6. asctime
  7. ño
  8. 跷
  9. <%=
  10. laşma

Suppresses

  1. list
  2. 列表
  3. list
  4. lists
  5. List
  6. List
  7. .list
  8. -list
  9. _list
  10. Lists