Neuron Explorer

Qwen3-0.6B

comma inside large numbers

A neuron in layer 8 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 the comma in big numbers like 1,526,006 or 1,332, the one that splits thousands from the rest. Turning it up made the model's text drift into ordinary sentences and numbers, which tells us little about the neuron itself.

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

    (250.2 km2)|⏎|• Water||0.7 sq mi (1.7 km2)|⏎|Elevation||1,148 ft (350 m)|⏎|Time zone||Eastern

  2. #2 peak

    As of the 2010 Census, the city had a population of 1,526,006, growing to 1,547

  3. #3 peak

    silver production were between 1872 and 1883. At the peak of its prosperity, the town had an estimated population of 6,000"⏎"

  4. #4 peak

    Ireland. The most notable jewel in the crown is a pale yellow brilliant in the centre of the front cross, weighing four carats. The crown contains 1,332 other diamonds, as well as pearls edging the band. Some

  5. #5 peak

    Mahadev Temple is located in Dhyansar village of Una district of Himachal Pradesh. It is situated at an altitude of about 1,500 meters above sea level and offers stunning views of the surrounding hills and

  6. #6 peak

    1, in World War II, the German battleship Bismarck sank the British battle cruiser Hood in the North Atlantic, with the loss of 1,300 lives.⏎In 1975 President Gerald Ford approved more

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: not, h. o. d. o. e. s. a. g. and. c. and. c. and. c. and. c. and. c. and. c.

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

    second voyage to the New World the French arrived off the coast of Florida on Thursday, June 22, 1564 around three or four o’clock in the afternoon. They landed about thirty leagues south of the St. Johns

  2. #2 peak

    nine. 3. My friend (to ring) me up at eight o’clock tomorrow. 4. Where Kate (to go) when you (to

  3. #3 peak

    it’s been planted. Watering slowly and deeply is better for your tree’s over-all health then fast and frequent waterings. Watering before 10 a.m. and after 8 p.m. can reduce the amount of water lost

  4. #4 peak

    few fires but the drought hit us hard.⏎In the hills surrounding the town and further afield the evidence was obvious with many stands of trees brown and dying.⏎Although this

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. L4:1657 excites
  2. L6:1299 suppresses
  3. L6:2016 excites
  4. L6:2363 suppresses
  5. L6:809 suppresses
  6. L7:2370 suppresses
  7. L6:184 excites
  8. L4:2827 excites
  9. L5:1387 excites
  10. L7:1684 excites
  11. L7:2139 excites
  12. L6:616 excites
  13. L5:2867 excites
  14. L7:2026 excites
  15. L6:2940 suppresses
  16. L6:1587 excites
  17. L7:2975 suppresses
  18. L6:785 excites
  19. L6:2938 suppresses
  20. L3:1661 suppresses
  21. L6:2264 excites
  22. L4:2974 suppresses
  23. L6:1695 excites
  24. L7:663 suppresses

Feeds later layers

  1. L15:646 strongly suppresses
  2. L16:592 strongly suppresses
  3. L12:576 suppresses 'of' in fixed phrases like table of contents
  4. L9:929 suppresses 'each' in 'each other'
  5. L17:1042 suppresses
  6. L9:209 suppresses
  7. L18:1321 suppresses
  8. L10:1705 suppresses
  9. L21:278 suppresses
  10. L16:2117 suppresses
  11. L9:116 suppresses
  12. L20:404 suppresses
  13. L19:107 excites
  14. L9:735 suppresses
  15. L16:286 suppresses
  16. L17:533 suppresses
  17. L10:678 suppresses
  18. L15:966 suppresses
  19. L11:300 suppresses
  20. L18:828 suppresses
  21. L25:645 suppresses
  22. L13:1552 suppresses
  23. L15:2051 suppresses
  24. L16:239 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 -4.69 and -4.491 values between -1.72 and -1.521 values between -1.52 and -1.322 values between -1.32 and -1.126 values between -1.12 and -0.926 values between -0.92 and -0.7316 values between -0.73 and -0.5357 values between -0.53 and -0.33468 values between -0.33 and -0.132364 values between -0.13 and 0.07854 values between 0.07 and 0.2789 values between 0.27 and 0.4619 values between 0.46 and 0.6612 values between 0.66 and 0.865 values between 0.86 and 1.062 values between 1.06 and 1.261 values between 1.26 and 1.462 values between 1.65 and 1.851 values between 1.85 and 2.051 values between 3.04 and 3.24offnegativepositiveoften

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. obil
  2. assigned
  3. returnValue
  4. Hayes
  5. perplex
  6. _w
  7. NF
  8. NV
  9. väl
  10. Michaels

Suppresses

  1. 场上
  2. 份
  3. ₋
  4. 霉素
  5. 荪
  6. 项
  7. iture
  8. ICIAL
  9. HandlerContext
  10. stance