Neuron Explorer

Qwen3-0.6B

unclear: chinese academic text, 'when' loops

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 mostly quiet and fires weakly on Chinese academic text and a few scattered words. Turning it down made the model repeat the word 'when' over and over.

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

    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。我们发现经过基于对比研究设计的汉语助动词课堂教学后

  2. #2 peak

    bs 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动词中

  3. #3 peak

    が含まれ、1人の参加者がいます。2020年7月6日 at 9:45 AMに Megan Smith さんが最後の更新を行いました。⏎2020年7月6日 9:4

  4. #4 peak

    densities."⏎The study, "Multi-State Trials of Bt Sweet Corn Varieties for Control of the Corn Earworm (Lepidoptera: Noctuidae)," analyzed the performance of Bt sweet corn, comparing its rate of inf

  5. #5 peak

    Hawaii's Oahu flattens from an island into a low-lying seamount. Researchers writing in the upcoming Feb. 15 issue of the journal Geochimica et Cosmochimica Acta estimate that the volcanic island will

  6. #6 peak

    plants, among them poison hemlock and fool's parsley ... Cow parsley can be confused with giant cow parsley/giant hogweed (Heracleum mantegazzianum), the sap of which can cause severe burns after coming in contact with 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: when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when when

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

    in the concentration of species in solution.⏎It’s frustrating that the article makes it seem like the applications of a really awesome discovery in pure science are mere moments away, rather than explaining why a discovery like this is awesome and encouraging peoples’ interest in

  2. #2 peak

    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动词中产生各种的偏误。本文通过调查法、统计

  3. #3 peak

    tendency, in some quarters, to jump to a solution like bariatric surgery, and rhapsodize about what an acceptable safe effective treatment alternative it is, without ever examining the underlying reasons for the obesity.⏎Reasons for obesity⏎Probably

  4. #4 peak

    Augmented Reality links knowledge/content with relevant visual imagery.⏎3 Steps to Creating an Awesome Virtual Museum in Class You're spending an afternoon browsing the exhibits at an art museum. If you're anything like me, you'd probably appreciate the art a lot

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. L18:1910 excites
  2. L19:219 suppresses
  3. L16:2782 excites
  4. L15:2140 excites
  5. L15:830 excites
  6. L15:1774 suppresses 'only' in 'not only ... but also'
  7. L19:1084 suppresses "and" joining clauses after a condition
  8. L16:1223 excites
  9. L18:845 weakly excites
  10. L18:409 weakly excites
  11. L18:1864 weakly excites
  12. L18:1249 weakly excites
  13. L18:1003 weakly excites
  14. L19:400 weakly excites
  15. L16:1556 weakly suppresses
  16. L18:1582 weakly suppresses
  17. L17:1081 weakly suppresses
  18. L18:251 weakly excites
  19. L18:2248 weakly excites
  20. L19:547 weakly excites
  21. L19:1966 weakly excites
  22. L0:193 weakly excites
  23. L12:758 weakly suppresses
  24. L15:2201 weakly excites

Feeds later layers

  1. L23:1645 strongly suppresses
  2. L27:659 strongly suppresses
  3. L25:633 strongly suppresses
  4. L23:449 strongly suppresses
  5. L21:1194 strongly suppresses
  6. L24:148 suppresses
  7. L22:676 suppresses
  8. L24:1542 excites
  9. L21:2960 suppresses
  10. L25:565 suppresses
  11. L25:1593 suppresses
  12. L22:1286 suppresses
  13. L23:2814 suppresses
  14. L24:2018 suppresses "first" in "for the first time"
  15. L21:1023 suppresses
  16. L26:1403 excites
  17. L22:1056 suppresses
  18. L26:2065 suppresses
  19. L26:814 suppresses
  20. L25:626 weakly suppresses
  21. L21:2942 weakly suppresses comma before a clause that contrasts a claim
  22. L21:1797 weakly suppresses
  23. L25:925 weakly suppresses
  24. L23:353 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 -1.74 and -1.661 values between -1.25 and -1.161 values between -1.16 and -1.081 values between -1.08 and -1.001 values between -1.00 and -0.923 values between -0.92 and -0.831 values between -0.67 and -0.591 values between -0.59 and -0.501 values between -0.50 and -0.426 values between -0.42 and -0.347 values between -0.34 and -0.2611 values between -0.26 and -0.1743 values between -0.17 and -0.09158 values between -0.09 and -0.01409 values between -0.01 and 0.07436 values between 0.07 and 0.16383 values between 0.16 and 0.24414 values between 0.24 and 0.32384 values between 0.32 and 0.40323 values between 0.40 and 0.49296 values between 0.49 and 0.57248 values between 0.57 and 0.65213 values between 0.65 and 0.73189 values between 0.73 and 0.82119 values between 0.82 and 0.9086 values between 0.90 and 0.9865 values between 0.98 and 1.0641 values between 1.06 and 1.1529 values between 1.15 and 1.2317 values between 1.23 and 1.3113 values between 1.31 and 1.393 values between 1.39 and 1.484 values between 1.48 and 1.56offnegativepositiveoften

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. 那么简单
  2. awei
  3. ENDIF
  4. rein
  5. 槿
  6. ayscale
  7. 衽
  8. ắng
  9. idences
  10. ложение

Suppresses

  1. 时
  2. 的时候
  3. when
  4. 时候
  5. when
  6. 時
  7. When
  8. 的時候
  9. 时表示
  10. When