Neuron Explorer

Qwen3-0.6B

opening parenthesis after a name or term

A neuron in layer 27 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 an opening bracket that follows a name, an abbreviation or a measurement, such as the short form that comes after a full organisation name or the metric value after an imperial one. Turning it up made the model write filler words like 'box' and 'sheet', so that link is weak.

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

    culicifacies, Bamisia tabaci, Cule<|endoftext|>The American Association of Feline Practitioners (AAFP) and the American Veterinary Medical Association (AVMA) discourage declawing.1,2 AAFP

  2. #2 peak

    edited for style and length.⏎Cite This Page:<|endoftext|>Briers, Mississippi facts for kids⏎Quick facts for kids⏎|Elevation||62 ft (19 m)|⏎|Time zone||UTC-6 (Central (C

  3. #3 peak

    umbu glacier at an elevation of 5300 m (17,400 ft); it receives an average of 450 mm (18 in) of precipitation a year. The climate of Mount Everest is extreme

  4. #4 peak

    was first petitioned to be listed under the Endangered Species Act (ESA) in 1995, and the U.S. Fish and Wildlife Service (USFWS) completed a 90-Day Finding of no warranted action in

  5. #5 peak

    shy animals rarely seen by visitors.⏎Until releases this week from Science Advances, it was believed there were two species of of wolves in the United States: gray wolves (Canis lupus) and red wolves (Canis rufus). Gray

  6. #6 peak

    have with their family, community, culture and place.⏎Being is about living here and now. Childhood is a special time in life and children need time to just ‘be’- time to play, try new things and have fun.⏎Bec

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: I need a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a Icreated a

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

    ?⏎If you are using open-source packages, then sooner or later, you will encounter the .tgz or .tar.gz files. This is a variation of zipped files; most open-source packages use this format to zip their files. If

  2. #2 peak

    will appear on the screen. Select Compress to Zip file.⏎Step 4: Wait for some time as the Zip file is processed. Once over, the zipped file will appear in the same folder as the original one.⏎Step 5

  3. #3 peak

    ? Birds that have the lower mandible crossing to the right are three times more common than those with a left-crossing lower mandible.⏎- White-winged Crossbills have been observed nesting and raising young across all 12 months of

  4. #4 peak

    .tar utility.⏎- x – Flag is used to extract the zipped files⏎- v – Flag is for verbose, or creating a list of

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. L25:222 strongly suppresses
  2. L19:1174 strongly excites
  3. L21:29 strongly excites
  4. L19:124 strongly suppresses
  5. L19:1207 strongly suppresses
  6. L25:239 strongly excites
  7. L20:563 strongly excites
  8. L22:65 strongly suppresses
  9. L20:1585 strongly excites
  10. L25:17 strongly excites
  11. L21:321 strongly excites
  12. L24:51 strongly excites
  13. L24:77 strongly excites opening parenthesis after a name
  14. L14:140 strongly suppresses
  15. L20:433 suppresses
  16. L23:262 suppresses
  17. L19:509 excites
  18. L19:1377 excites
  19. L18:76 suppresses blank spot before a number
  20. L20:2882 suppresses
  21. L22:216 suppresses
  22. L23:90 suppresses
  23. L23:246 excites
  24. L20:193 excites

Feeds later layers

None: this is the last layer.

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 -44.41 and -42.541 values between -38.81 and -36.951 values between -35.08 and -33.221 values between -29.49 and -27.632 values between -27.63 and -25.766 values between -25.76 and -23.902 values between -22.03 and -20.175 values between -20.17 and -18.301 values between -18.30 and -16.448 values between -16.44 and -14.5810 values between -14.58 and -12.719 values between -12.71 and -10.8516 values between -10.85 and -8.9812 values between -8.98 and -7.1224 values between -7.12 and -5.2533 values between -5.25 and -3.3950 values between -3.39 and -1.523206 values between -1.52 and 0.34359 values between 0.34 and 2.2157 values between 2.21 and 4.0741 values between 4.07 and 5.9326 values between 5.93 and 7.8011 values between 7.80 and 9.667 values between 9.66 and 11.535 values between 11.53 and 13.395 values between 13.39 and 15.262 values between 15.26 and 17.123 values between 17.12 and 18.991 values between 22.71 and 24.581 values between 24.58 and 26.442 values between 28.31 and 30.17offnegativepositiveoften

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. box
  2. sheet
  3. ListNode
  4. 词
  5. logging
  6. fs
  7. -sheet
  8. db
  9. ListModel
  10. 纲

Suppresses

  1. created
  2. soever
  3. adays
  4. Again
  5. outside
  6. again
  7. through
  8. above
  9. 还不是
  10. after