Neuron Explorer

Qwen3-0.6B

'they' or 'we' after where or who

A neuron in layer 26 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 fires on a pronoun like "they" or "we" that comes right after words like "where", "who" or "if", as in "who we are" or "where they started". Turned up, the output drifted into unrelated text about news and reflections.

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

    positive and negative, which sum to zero.⏎The positive cycle may move charge carriers, such as electrons. But then the negative cycle pulls the charges back to where they started. To reliably control the quantum information, an asymmetric light wave is needed.⏎

  2. #2 peak

    the starting point illustrated in their body rotation. Most of the participants could use information about where they were in the room just a few moments before to tell them where they were in relation to the starting point.⏎Further research into where the “noise”

  3. #3 peak

    included this in Genesis 3<|endoftext|>Sujata Gupta for Nature – Brain food: Clever eating:⏎Our ancestors’ consumption of meat is what made us who we are.⏎

  4. #4 peak

    continue building out ground stations as it gets more approvals from the regulatory bodies to do so in the United States and Canada.<|endoftext|>Our brain is what makes us who we are, is it not? All aspects of our life are stored into this one

  5. #5 peak

    pointier and whiter. So the size and other teeth features should help you determine its age.⏎Also, check closely if the permanent teeth are white. If they still are, then your cat is probably around 16

  6. #6 peak

    chalices, and other spoils of the Church might be as much more. These movables were before night taken to the King’s treasury—a sign that they will not be restored.”1⏎The Spanish Chronicle also reported Thomas Cromwell’s

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: howbeitbeitartreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeitreambeit

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

    evaluations about jesuitic missions among the Guaranis - Relação abbreviada da republica, que os religiosos jesuitas estabelecerão no Paraguai, by Marquis of Pombal, Candide

  2. #2 peak

    a masculine, plural noun so the masculine plural direct object pronoun los is used to replace it in the second sentence.Son muy buenos hijos. No los cambiaría.⏎They are good

  3. #3 peak

    (Bot.) A smooth-skinned variety of peach. [1913 Webster] Spanish nectarine, the plumlike fruit of the West Indian tree Chrysobalanus Icaco; -- also called cocoa plum. It is

  4. #4 peak

    2018, Schuldt et al. 2020). Under these circumstances, tree´s survival

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. L23:625 strongly suppresses
  2. L22:309 strongly excites
  3. L24:120 strongly suppresses
  4. L24:241 strongly suppresses
  5. L23:176 strongly suppresses
  6. L24:600 strongly excites
  7. L20:1423 strongly suppresses
  8. L23:287 strongly suppresses 'are' after a plural subject
  9. L21:468 strongly excites
  10. L24:2880 strongly excites
  11. L19:1317 strongly excites
  12. L25:2055 strongly suppresses
  13. L25:2082 strongly suppresses first-person 'I' starting a clause
  14. L23:1210 strongly suppresses
  15. L19:1422 strongly excites
  16. L24:401 strongly suppresses 'it' referring back to a noun
  17. L21:2936 strongly suppresses
  18. L18:444 suppresses
  19. L24:2161 excites
  20. L20:2453 excites blank slot before a date
  21. L12:642 excites
  22. L25:586 suppresses
  23. L16:1317 excites
  24. L2:149 suppresses

Feeds later layers

  1. L27:500 strongly suppresses
  2. L27:1256 strongly suppresses
  3. L27:321 suppresses
  4. L27:650 excites
  5. L27:461 suppresses
  6. L27:1040 suppresses
  7. L27:804 weakly suppresses
  8. L27:2212 weakly suppresses
  9. L27:2329 weakly suppresses
  10. L27:1271 weakly suppresses
  11. L27:1799 weakly suppresses
  12. L27:3047 weakly suppresses
  13. L27:1911 weakly suppresses
  14. L27:168 weakly suppresses
  15. L27:995 weakly suppresses
  16. L27:2196 weakly suppresses
  17. L27:1740 weakly suppresses
  18. L27:3032 weakly suppresses 're' prefix meaning again
  19. L27:485 weakly suppresses
  20. L27:1357 weakly excites
  21. L27:1937 weakly suppresses
  22. L27:2884 weakly suppresses
  23. L27:2900 weakly suppresses
  24. L27:928 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 -7.66 and -5.943 values between -5.94 and -4.215 values between -4.21 and -2.49610 values between -2.49 and -0.761814 values between -0.76 and 0.96621 values between 0.96 and 2.69324 values between 2.69 and 4.41190 values between 4.41 and 6.1497 values between 6.14 and 7.8652 values between 7.86 and 9.5940 values between 9.59 and 11.3125 values between 11.31 and 13.0416 values between 13.04 and 14.7613 values between 14.76 and 16.4910 values between 16.49 and 18.219 values between 18.21 and 19.946 values between 19.94 and 21.666 values between 21.66 and 23.398 values between 23.39 and 25.124 values between 25.12 and 26.846 values between 26.84 and 28.576 values between 28.57 and 30.296 values between 30.29 and 32.023 values between 32.02 and 33.742 values between 33.74 and 35.479 values between 35.47 and 37.194 values between 37.19 and 38.926 values between 38.92 and 40.643 values between 40.64 and 42.371 values between 42.37 and 44.093 values between 44.09 and 45.821 values between 47.54 and 49.272 values between 49.27 and 50.991 values between 52.72 and 54.441 values between 59.62 and 61.34offnegativepositiveoften

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. 都要
  2. 都会
  3. 都能
  4. 都需要
  5. 都会有
  6. 都能够
  7. 都有着
  8. 都说
  9. 都非常
  10. 都不会

Suppresses

  1. was
  2. is
  3. was
  4. wasn
  5. isn
  6. 的是
  7. $is
  8. →is
  9. (is
  10. "is