Neuron Explorer

Qwen3-0.6B

comma after a speaker tag

A neuron in layer 18 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 responds to a comma that comes right after a person's name or a "said" phrase, such as "said Markram," or "Halloran said,", where a quote or a clause is about to continue. It is pushed off by commas in long, formal old-style sentences. Turning it up gave wordy, rambling text without a clear theme.

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

    create models to explore how neural circuits are organized, and how they give rise to behaviour and cognition — among the deepest mysteries in neuroscience. Ultimately, said Markram, the HBP would even help researchers to grapple with disorders such as Alzheimer’s

  2. #2 peak

    .⏎The Cherokee presented an interesting dilemma. It was declared, with a sigh, that most Indians would be removed due merely to the<|endoftext|>Editorial: Team effort

  3. #3 peak

    ranging from yeast to fruit flies to mice to humans. The idea, said study researcher Donard Dwyer, is that in people with a broken starvation response, a few initial rounds of diet

  4. #4 peak

    be compared to a wise man who built his house on the rock……⏎Revelation 3:20 'Behold, I stand at the door and knock; if anyone hears My voice and opens the

  5. #5 peak

    .”⏎Not surprisingly, notes Lo, as the dose of THC increased, so did the impact to menstrual cycle length and

  6. #6 peak

    protect against enterovirus infection, which can cause high fever, diarrhea and sometimes death.⏎Similarly, they say, there is a strong correlation between rheumatoid arthritis and tuberculosis. Populations with

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 have, as, or to be, as, or, as, or, as, or, as, or, as, or, as, or, as, or, as, or, as, or, as, or

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

    III.) a terrible pestilence devastated London. The dirt and crowding of the old mediæval cities made them at all times nurseries of infectious disease, and when a great

  2. #2 peak

    among savages than among civilised men. Notwithstanding the differences which necessarily exist between savage and civilised ethics, the two systems must evidently have much in common, for from the days of Columbus to the present, travellers have given testimony of customs

  3. #3 peak

    effected by Themistocles who had meanwhile created the Athenian fleet which at Salamis proved its superiority over the Perso-Phoenician armada, anc thus precluded beforehand the success of the land-forces.<|endoftext|>The Treasure

  4. #4 peak

    uary will receive from guests more attention than will be accorded to these ancient weapons.⏎No doubt if one could procure a shell strung with gold and silver cords, after the fashion of the old-time instrument with which the gods made music, the

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. L10:315 excites
  2. L4:90 weakly suppresses
  3. L3:1758 weakly excites
  4. L10:2217 weakly suppresses
  5. L15:995 weakly suppresses
  6. L12:35 weakly suppresses
  7. L1:1007 weakly excites name pieces inside a multi-part word
  8. L14:1832 weakly excites
  9. L7:2925 weakly suppresses
  10. L13:44 weakly excites inside web addresses and citations
  11. L15:2372 weakly suppresses
  12. L5:2342 weakly suppresses
  13. L10:1042 weakly suppresses
  14. L12:1968 weakly suppresses
  15. L15:930 weakly suppresses
  16. L16:245 weakly suppresses
  17. L2:1207 weakly suppresses
  18. L16:775 weakly excites
  19. L16:1529 weakly excites
  20. L1:2910 weakly suppresses
  21. L8:1400 weakly excites
  22. L4:726 weakly excites
  23. L8:2822 weakly excites
  24. L4:76 weakly suppresses

Feeds later layers

  1. L21:344 excites
  2. L19:625 weakly suppresses
  3. L19:1034 weakly suppresses
  4. L22:368 weakly suppresses
  5. L19:1172 weakly excites
  6. L19:1414 weakly suppresses
  7. L21:632 weakly suppresses
  8. L19:1310 weakly excites
  9. L22:158 weakly excites
  10. L26:462 weakly suppresses
  11. L23:1145 weakly excites
  12. L25:308 weakly suppresses
  13. L20:116 weakly excites
  14. L19:1803 weakly excites
  15. L21:906 weakly excites
  16. L21:2161 weakly excites
  17. L20:596 weakly excites
  18. L19:482 weakly excites
  19. L20:1391 weakly suppresses
  20. L22:786 weakly suppresses
  21. L20:2133 weakly suppresses letters inside spelled-out words
  22. L22:1024 weakly excites
  23. L21:2019 weakly suppresses
  24. L19:1558 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 -10.94 and -10.421 values between -9.89 and -9.371 values between -9.37 and -8.851 values between -8.85 and -8.331 values between -8.33 and -7.812 values between -7.28 and -6.764 values between -6.76 and -6.243 values between -6.24 and -5.721 values between -5.72 and -5.2014 values between -5.20 and -4.678 values between -4.67 and -4.1522 values between -4.15 and -3.6317 values between -3.63 and -3.1154 values between -3.11 and -2.5876 values between -2.58 and -2.06123 values between -2.06 and -1.54207 values between -1.54 and -1.02302 values between -1.02 and -0.50816 values between -0.50 and 0.03939 values between 0.03 and 0.55566 values between 0.55 and 1.07351 values between 1.07 and 1.59180 values between 1.59 and 2.11107 values between 2.11 and 2.6462 values between 2.64 and 3.1627 values between 3.16 and 3.6813 values between 3.68 and 4.203 values between 4.20 and 4.732 values between 4.73 and 5.251 values between 5.25 and 5.772 values between 5.77 and 6.291 values between 9.42 and 9.95offnegativepositiveoften

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. REAT
  2. yy
  3. 的一款
  4. igor
  5. -word
  6. 心
  7. 你就
  8. ị
  9. نفس
  10. odon

Suppresses

  1. 这批
  2. 起来了
  3. ठ
  4. брос
  5. gemäß
  6. udence
  7. etti
  8. dings
  9. .social
  10. etty