Neuron Explorer

Qwen3-0.6B

comma before 'right?' tag

A neuron in layer 24 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

Fires on a comma that comes just before a casual tag like 'right?' or 'does not it!', where the writer checks agreement with the reader. Turning it up made the model repeat the word 'right' over and over.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    side of the moon. Its mission? To take pictures of this unexplored part of the moon, collect soil samples, and plant a mini garden. Awesome, right? So how successful has the Chang’e

  2. #2 peak

    . Nothing travels faster than the speed of light, but even light takes time to get to places. It takes a ray of sunlight

  3. #3 peak

    it.⏎Anyway, what about the link between a woman’s menstrual cycle and the lunar cycle? I mean, that’s why we’re all still reading this post, right?⏎Well, let’s start off with a bit of backstory: For generations

  4. #4 peak

    learning areas is simpler for you to collaborate with, and will certainly use up much less of your time than searching for individual worksheets.⏎This sounds like a horrible lot, does not it! A great collection of preschool worksheets should cover every one of this

  5. #5 peak

    and LeBron James refer to themselves by their own name, it's called illeism, and it can actually be a useful psychological technique.⏎Recently, Grammar Girl listener Mark

  6. #6 peak

    : ‘do not’ would change to ‘don’t’⏎That’s it, you’ve got it…. Now write these sentences again but add an apostrophe in

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: how to best serve the needs of a school can be considered as a critical reappearance of the national can be considered as a critical reappearance of the national? - - - - - - - - - - - - - - - -

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

    of the countless missing links that should have numbered in the millions.⏎J. L. B. Smith, posing with the second coelacanth caught off the Comora Islands in 1952.⏎This coelacanth fossil,

  2. #2 peak

    0. It connected the San Miguel district to Malate and Ermita.⏎The bridge also served Isla de Convalescencia, the island in the Pasig River where the orphanage of Hospicio de San Jose is located.⏎Information

  3. #3 peak

    at Madrid, 27 March, 1625. He was a great-grandson of the Tordesillas who was put to death by the Comuneros at Seville. He studied in Spain and Italy, and became secretary

  4. #4 peak

    1254), as well as hospitals at Pontoise, Vernon, and Compiégne.⏎St. Louis was a patron of architecture. The Sainte

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. L20:2335 excites comma ending an 'if' clause
  2. L17:3046 suppresses answer labels and tag questions
  3. L21:1538 suppresses
  4. L20:1371 suppresses comma inside a question
  5. L23:1431 excites
  6. L19:650 weakly suppresses
  7. L15:271 weakly excites
  8. L11:1624 weakly excites
  9. L7:1064 weakly excites
  10. L20:1922 weakly suppresses
  11. L17:2456 weakly suppresses
  12. L12:731 weakly excites
  13. L20:1453 weakly suppresses
  14. L23:1028 weakly suppresses
  15. L22:697 weakly excites
  16. L20:2490 weakly excites
  17. L12:1237 weakly excites
  18. L19:1036 weakly suppresses
  19. L8:2204 weakly suppresses
  20. L2:231 weakly suppresses
  21. L19:1274 weakly suppresses
  22. L13:2326 weakly excites
  23. L2:2206 weakly suppresses
  24. L13:335 weakly suppresses

Feeds later layers

  1. L27:1296 strongly excites
  2. L26:1121 excites
  3. L27:1489 excites
  4. L27:1760 excites
  5. L25:1754 excites
  6. L26:930 weakly suppresses
  7. L27:1592 weakly excites
  8. L25:373 weakly excites
  9. L25:1843 weakly excites
  10. L25:1156 weakly excites
  11. L26:406 weakly excites
  12. L27:1212 weakly suppresses
  13. L27:302 weakly excites
  14. L26:2397 weakly excites
  15. L27:2800 weakly excites
  16. L25:651 weakly excites
  17. L27:2914 weakly suppresses
  18. L25:2798 weakly suppresses
  19. L25:2067 weakly suppresses
  20. L25:1656 weakly excites
  21. L27:1261 weakly excites
  22. L25:2211 weakly excites
  23. L26:1896 weakly suppresses
  24. L26:1820 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 -18.27 and -17.301 values between -10.51 and -9.541 values between -9.54 and -8.571 values between -7.60 and -6.631 values between -6.63 and -5.663 values between -5.66 and -4.691 values between -3.72 and -2.758 values between -2.75 and -1.7876 values between -1.78 and -0.813065 values between -0.81 and 0.16669 values between 0.16 and 1.1329 values between 1.13 and 2.1013 values between 2.10 and 3.0712 values between 3.07 and 4.048 values between 4.04 and 5.012 values between 5.01 and 5.983 values between 5.98 and 6.953 values between 6.95 and 7.924 values between 7.92 and 8.893 values between 8.89 and 9.861 values between 9.86 and 10.831 values between 10.83 and 11.801 values between 12.77 and 13.741 values between 19.56 and 20.53offnegativepositiveoften

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. right
  2. right
  3. Right
  4. Right
  5. →right
  6. (right
  7. RIGHT
  8. 右
  9. -right
  10. ,right

Suppresses

  1. 呢
  2. ivr
  3. 嗎
  4. 망
  5. 吗
  6. 厚度
  7. moz
  8. �数
  9. 低迷
  10. .GetInt