Neuron Explorer

Qwen3-0.6B

chinese 的 particle before a noun

A neuron in layer 0 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 mostly fires on the Chinese word 的 (and similar linking words) when it comes before a noun, especially in Chinese academic text. It is much weaker on English, where it fires on a few words like 'insidious'. Turning it up or down did not give a clear change in the model's English 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

    法、分析法来进行研究,探讨泰国学生在学习汉语助动词中的困难及偏误情况,分析他们产生偏误的原因。我们发现经过

  2. #2 peak

    Chinese Auxiliary Verbs for Thai Learners Based on the Comparative Analysis⏎- 人文学院-已发表论文 ⏎“助动词”在汉语中具有较大的作用,使用率高,但是泰国学生经常在学习和使用汉语助

  3. #3 peak

    to describe a global TCP/IP network. So, you could also argue the Internet (<|endoftext|>このトピックには0件の返信が含まれ、1人の参加者がいます。2020年7月6日 at 9

  4. #4 peak

    , and procrastinating until the last moment can feel overwhelming to both students and parents. Additionally, it

  5. #5 peak

    ��性病的⏎anachronism /əˈnækrəˌnɪz

  6. #6 peak

    If you've never experienced the monster that is depression, there is really no way for you to grasp its horrors. One of the things that makes depression so insidious is that it distorts one's self-perception and causes the belief that the

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 want to be more involved in the world, and I want to be more involved in the world, and I want to be more involved in the world, and I want to be more involved in the world, and I want to be

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 paying back on time that you can build up a good score.”⏎Happily, while In addition, three in ten people are worried<|endoftext|>Formation Of Plural Nouns⏎An English noun exhibits a two-way distinction of number: a singular

  2. #2 peak

    ten objects from museum collections in Greater Manchester⏎A compass used in a PoW escape, a railway station sundial and part of the world's first commercial computer: objects that tell a history of

  3. #3 peak

    ist (pointillist) techniques.⏎He also studied with Henri de Toulouse-Lautrec, and he exhibited works at Cafe Tambourin with Louis Anquetin, Emile Bernard, and Toulouse-Laut

  4. #4 peak

    they can get advantages anytime.⏎Why Expertsmind for assignment help⏎- Higher degree holder and experienced experts network⏎- Punctuality and responsibility of work⏎- Quality solution with 100% plagiarism

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

None: this is the first layer.

Feeds later layers

  1. L4:300 strongly excites
  2. L2:1473 excites
  3. L12:1058 excites
  4. L16:1433 excites
  5. L3:1774 excites
  6. L6:1063 excites
  7. L17:3043 excites
  8. L7:111 excites
  9. L18:1265 excites
  10. L6:1233 excites
  11. L26:1776 excites
  12. L15:286 excites
  13. L21:2486 excites
  14. L9:366 excites
  15. L9:240 excites
  16. L8:2755 excites
  17. L4:2743 excites
  18. L5:1921 excites
  19. L8:252 suppresses 'of' after 'each' or 'one' or 'all'
  20. L18:602 excites
  21. L13:1196 excites opening paren before an acronym
  22. L14:118 suppresses
  23. L14:280 weakly suppresses
  24. L7:210 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.

1 values between -0.28 and -0.261 values between -0.26 and -0.243 values between -0.22 and -0.203 values between -0.20 and -0.194 values between -0.19 and -0.1715 values between -0.17 and -0.1531 values between -0.15 and -0.1355 values between -0.13 and -0.1192 values between -0.11 and -0.09144 values between -0.09 and -0.07228 values between -0.07 and -0.05283 values between -0.05 and -0.03404 values between -0.03 and -0.01572 values between -0.01 and 0.01474 values between 0.01 and 0.03431 values between 0.03 and 0.05317 values between 0.05 and 0.07294 values between 0.07 and 0.09190 values between 0.09 and 0.11155 values between 0.11 and 0.13105 values between 0.13 and 0.1459 values between 0.14 and 0.1628 values between 0.16 and 0.1810 values between 0.18 and 0.204 values between 0.20 and 0.222 values between 0.22 and 0.241 values between 0.24 and 0.261 values between 0.32 and 0.341 values between 0.47 and 0.49offnegativepositiveoften

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. urga
  3. `,
  4. uddenly
  5. 郎
  6. loạt
  7. ⏎ ⏎
  8. `,⏎
  9. бег
  10. URITY

Suppresses

  1. _Reference
  2. enschaft
  3. 的真实性
  4. ance
  5. 性的
  6. raising
  7. 一楼
  8. aticon
  9. unas
  10. 人民服务