Neuron Explorer

Qwen3-0.6B

comma after a clause's verb or subject

A neuron in layer 22 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 comma that comes right after a verb or a subject, where the sentence is about to continue with a new clause, such as "and, later" or "who, while she lived". Turned up, the model produced repeated "with" phrases, and turned down it produced strange word chains.

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

    3200 BC, when there was a small settlement on the bank of the Nile where Karnak now stands. The great temple complex at Karnak is, however, mostly a Middle Kingdom creation. Archaeological excavation reveals that the complex was

  2. #2 peak

    the people of Novgorod to rule that city (c. 862); the Rurik princes maintained their control over Kievan Rus and, later, Muscovy until 1598.⏎Rurik’s

  3. #3 peak

    and the wife of a future prime minister of the United Kingdom, Primrose was in her day the wealthiest woman in Britain, and an astute political observer who, while she lived, played a key role in advancing her husband’s career.⏎Like

  4. #4 peak

    Hebrides that Columba turned his face when he accepted the terrible penance of Molaise. He bade farewell to his relatives, and, with a few monks who insisted on accompany him whithersoever he might

  5. #5 peak

    amusement.⏎There was a half-decayed log lying by the roadside, and Grant sat down on it, calmly pulled a cigar from his cigar case, and, requesting one of his staff who stood nearby to furnish him a sheet of paper,

  6. #6 peak

    interests would be better served, he continued, if Japan instead of Russia eventually dominated Korea.⏎Kennan urged the secretary to reign in General Douglas MacArthur, who, he wrote, had “wide and relatively uncontrolled latitude… in determining our policy

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: to the left of the initial then the then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then then

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

    has allowed me to have fun in life, to go to school, and to make a choice in life. I can watch a movie, or I don’t have to. I could hang out with my friends or I could hang out with

  2. #2 peak

    and answering the questions using the text.⏎Please read the sheet below ‘All About Lighthouses’ and then answer the questions below. Please remember, your answers need to be in full sentences and using your cursive handwriting and correct punctuation (cap

  3. #3 peak

    felt animals are chosen, another list would drop down showing the bag with 50 animals and then, the bag with 100 animals. The same can be done for handmade coats. You might

  4. #4 peak

    toward protecting family members from the risks of fire. This includes installing smoke detectors, keeping fire extinguishers charge, and establishing an emergency plan in case family members need to exit the house quickly if a fire starts.⏎However, fewer people know how

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. L18:572 weakly suppresses
  2. L7:92 weakly suppresses
  3. L21:456 weakly suppresses
  4. L20:2977 weakly suppresses
  5. L16:1059 weakly excites
  6. L21:583 weakly suppresses
  7. L18:2476 weakly suppresses
  8. L21:2516 weakly suppresses
  9. L16:941 weakly suppresses
  10. L13:685 weakly suppresses
  11. L2:629 weakly suppresses
  12. L21:109 weakly excites
  13. L14:2733 weakly suppresses
  14. L7:1510 weakly suppresses
  15. L15:184 weakly suppresses
  16. L11:840 weakly excites
  17. L16:3001 weakly suppresses
  18. L17:2336 weakly suppresses
  19. L11:223 weakly suppresses
  20. L10:873 weakly suppresses
  21. L17:2831 weakly suppresses
  22. L19:1306 weakly excites
  23. L6:1280 weakly suppresses
  24. L21:1339 weakly excites

Feeds later layers

  1. L24:148 excites
  2. L27:603 excites
  3. L23:984 excites
  4. L24:116 suppresses
  5. L26:852 excites
  6. L23:756 excites
  7. L25:806 excites
  8. L24:883 excites
  9. L23:664 excites
  10. L26:741 weakly excites
  11. L26:2350 weakly excites
  12. L23:326 weakly excites
  13. L26:708 weakly excites
  14. L24:396 weakly excites
  15. L23:1590 weakly excites
  16. L24:921 weakly excites
  17. L25:1186 weakly excites
  18. L24:1173 weakly excites
  19. L24:1690 weakly excites
  20. L24:784 weakly excites
  21. L23:570 weakly excites
  22. L24:492 weakly excites
  23. L24:1711 weakly excites
  24. L27:927 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 -21.03 and -19.541 values between -18.06 and -16.573 values between -16.57 and -15.081 values between -15.08 and -13.597 values between -13.59 and -12.117 values between -12.11 and -10.6213 values between -10.62 and -9.1317 values between -9.13 and -7.6450 values between -7.64 and -6.1691 values between -6.16 and -4.67148 values between -4.67 and -3.18319 values between -3.18 and -1.69845 values between -1.69 and -0.212026 values between -0.21 and 1.28175 values between 1.28 and 2.7786 values between 2.77 and 4.2654 values between 4.26 and 5.7427 values between 5.74 and 7.2313 values between 7.23 and 8.724 values between 8.72 and 10.219 values between 10.21 and 11.693 values between 11.69 and 13.182 values between 13.18 and 14.671 values between 19.13 and 20.621 values between 20.62 and 22.112 values between 23.59 and 25.081 values between 26.57 and 28.061 values between 36.98 and 38.47offnegativepositiveoften

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. swith
  2. with
  3. without
  4. with
  5. WITH
  6. avec
  7. -with
  8. without
  9. through
  10. With

Suppresses

  1. JSGlobalScope
  2. 他又
  3. entonces
  4. 他曾
  5. 時候
  6. 我当时
  7. 他是
  8. then
  9. 骢
  10. amient