Neuron Explorer

Qwen3-0.6B

comma before a new clause

A neuron in layer 21 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 lights up on a comma that comes right after a complete statement, just before the writer starts a new clause such as 'and the second is...' or 'but it...'. When it was turned up, the model fell into long loops of 'and, and, and'.

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

    advantages over using a single drive to save data - the first one is redundancy, so if one drive breaks down, the data will be accessible from the remaining ones, and the second is improved performance as the input/output, or reading/writing operations

  2. #2 peak

    24 and eighth-graders scored 274. The national average for fourth- and eighth-graders was 217 and 263, respectively.⏎Only Connecticut, Massachusetts, Montana and North Dakota scored higher than Vermont’s

  3. #3 peak

    between the two have been the source of conflict for thousands of years.⏎Both the Jewish and Muslim faith believe in one God. Being monotheistic means just that, the belief in one God. This is the most obvious similarity between the two religions

  4. #4 peak

    , and Reinforcement Learning.⏎Start with the question, “Do you have a Target Variable?” If the answer is ‘Yes’, it is a supervised learning problem, and if the answer is ‘No’, it

  5. #5 peak

    AC⏎An airport in Japan will start collecting snow in 2010 to keep the terminals cool during the hot summer months. It sounds a bit crazy, but the small agricultural community of Funagata has been using snow technology since

  6. #6 peak

    two main blood groups, the ABO group and the Rh group. The ABO group is made up of A, B, AB, or O blood types, and the Rh group is either positive or negative. You blood groups help doctors know

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: the current status of the political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political political

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

    a series of images of the same blurry face — with the hint that it was a non-American political leader — as it would appear from 200, 100 and 50 meters. At 50 meters,

  2. #2 peak

    litres of water a day, the UN Food and Agriculture Organisation (FAO) calculates that the water needed to produce a day’s food can range between 2,000 and 5,000 litres.⏎It is hard to

  3. #3 peak

    ⏎Here’s another great example video done by Nick in After Effects using precomps!⏎Check back next week to learn the difference between frame rate and shutter speed, or view more lessons at the Vimeo Video School .<|endoftext|>Some time back I presented

  4. #4 peak

    00 DPI optical mice as well as 800 DPI laser mice are very popular, but some high-end models are capable of 1000, 1600 or even 2000 DPI tracking speeds.⏎

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. L19:682 strongly excites
  2. L20:1358 excites
  3. L18:366 excites comma inside numbers and dates
  4. L18:1394 excites
  5. L20:2082 excites
  6. L17:116 excites
  7. L16:240 excites comma after 'sometimes' or 'the truth'
  8. L14:35 suppresses
  9. L19:726 suppresses comma inside large numbers
  10. L18:621 suppresses
  11. L20:610 suppresses
  12. L18:174 suppresses
  13. L17:118 excites comma after a clause ending
  14. L20:1744 suppresses comma after a sentence's subject
  15. L19:2451 excites
  16. L17:1405 suppresses
  17. L16:1082 suppresses comma linking two clauses after 'and'
  18. L9:77 excites commas in citations and name lists
  19. L18:2376 suppresses
  20. L18:2593 suppresses
  21. L19:2229 suppresses
  22. L18:1791 excites
  23. L16:241 excites
  24. L20:2730 suppresses

Feeds later layers

  1. L22:368 strongly excites
  2. L27:86 strongly suppresses
  3. L26:22 strongly suppresses the 'res' fragment inside words
  4. L27:358 strongly excites
  5. L26:236 strongly excites
  6. L25:328 strongly excites
  7. L22:158 strongly suppresses
  8. L27:130 suppresses
  9. L27:191 excites
  10. L27:286 excites
  11. L25:1221 excites
  12. L24:32 excites
  13. L26:80 suppresses
  14. L26:345 excites
  15. L25:438 excites
  16. L26:844 suppresses
  17. L24:216 excites
  18. L26:307 suppresses
  19. L24:199 suppresses
  20. L26:456 excites
  21. L27:77 suppresses
  22. L25:381 suppresses
  23. L22:1262 suppresses
  24. L26:1212 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 -13.18 and -12.551 values between -8.15 and -7.521 values between -7.52 and -6.892 values between -6.89 and -6.265 values between -6.26 and -5.631 values between -5.63 and -5.006 values between -5.00 and -4.389 values between -4.38 and -3.756 values between -3.75 and -3.1210 values between -3.12 and -2.4912 values between -2.49 and -1.8623 values between -1.86 and -1.2345 values between -1.23 and -0.602018 values between -0.60 and 0.031550 values between 0.03 and 0.66130 values between 0.66 and 1.2936 values between 1.29 and 1.9115 values between 1.91 and 2.548 values between 2.54 and 3.176 values between 3.17 and 3.804 values between 3.80 and 4.435 values between 4.43 and 5.063 values between 5.06 and 5.694 values between 5.69 and 6.324 values between 6.95 and 7.571 values between 7.57 and 8.201 values between 8.83 and 9.461 values between 11.35 and 11.98offnegativepositiveoften

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. ,
  2. ,',
  3. ,...⏎⏎
  4. ,en
  5. ,C
  6. ,q
  7. ,default
  8. ,rp
  9. ,m
  10. ,W

Suppresses

  1. ||⏎
  2. 善意
  3. lay
  4. しました
  5. "]["
  6. しておく
  7. ].[
  8. emsp
  9. OwnProperty
  10. ночь