Neuron Explorer

Qwen3-0.6B

zero inside a number

A neuron in layer 10 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 '0' that sits inside a number, like 50, 40 or 1,036, in tables, statistics and measurements. Turning it up made the model write unrelated, rambling text.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    ||190 beats per minute|⏎|35 years||93-157 beats per minute||185 beats per minute|⏎|40 years||90-153 beats per minute||180 beats

  2. #2 peak

    (49.6% girls; mean age, 13.5 years), 582 in 10th grade (50.0% girls; mean age, 15.4 years), and 5

  3. #3 peak

    vised)|⏎|Displacement:||925 long tons (940 t; 1,036 short tons)|⏎|Length:||205 ft (6

  4. #4 peak

    5% CI, 1.7-1.9) for 2009 to 2012 to an RR of 3.0 (95% CI

  5. #5 peak

    McKinley, North America⏎At the Denali National Park in the US state of Alaska lies the 20,320 ft (6,194 m) Mount McKinley or Denali (its Koyukon

  6. #6 peak

    have walked on land. These include the mighty Argentinosaurus and Alamosaurus, thought to have tipped the scales at around 73 tonnes (80 tons) each. Despite not quite attaining those sizes, the new species stands

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: If I have a lot of time, and if I have a lot of time, if I have a lot of time, if I have a lot of time, if I have a lot of time, if I have a lot of time

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

    meditation on the meaning of the story of Lot's wife.⏎This article is available to subscribers only. Please subscribe for full access—subscriptions begin at $2.95. Already have an online account? Log in now. Already a print subscriber

  2. #2 peak

    to three different classmates. (No fewer than 100 words). Please remember this post and your replies to classmates are due Thursday by 11:59 pm. You are required to cite material from your reading in your post.

  3. #3 peak

    ;⏎P0 = 0x80;⏎P0 = 0x98;⏎P0 = 0xC0;⏎P0 = 0xff;⏎} while(1); // Endless loop<|endoftext|>Wisconsin Well Water:

  4. #4 peak

    :59 PM MDT · Views: 1,130⏎HUMAN E

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. L9:1032 excites
  2. L5:2950 weakly excites
  3. L3:2928 weakly suppresses
  4. L4:201 weakly excites
  5. L9:2379 weakly excites
  6. L0:1400 weakly suppresses
  7. L4:2878 weakly suppresses
  8. L2:2043 weakly excites
  9. L9:316 weakly excites
  10. L1:1070 weakly suppresses
  11. L5:2359 weakly excites
  12. L7:780 weakly suppresses
  13. L7:2126 weakly suppresses
  14. L9:802 weakly excites
  15. L7:2423 weakly suppresses
  16. L9:1326 weakly suppresses
  17. L7:2837 weakly excites
  18. L9:1957 weakly excites
  19. L8:1086 weakly excites
  20. L8:2027 weakly suppresses
  21. L1:1862 weakly excites
  22. L3:2426 weakly suppresses
  23. L6:1454 weakly excites
  24. L7:2448 weakly excites

Feeds later layers

  1. L11:560 suppresses
  2. L11:689 suppresses
  3. L14:2654 weakly excites
  4. L14:2834 weakly suppresses
  5. L19:2311 weakly excites
  6. L23:1406 weakly excites
  7. L20:2395 weakly excites
  8. L20:2611 weakly excites
  9. L23:920 weakly suppresses
  10. L16:3070 weakly excites
  11. L26:1259 weakly suppresses
  12. L26:2032 weakly excites 'to' or 'from' in 'from one X to another'
  13. L25:3059 weakly suppresses
  14. L19:843 weakly excites
  15. L23:1049 weakly excites
  16. L20:3007 weakly excites
  17. L19:1679 weakly suppresses
  18. L23:1648 weakly suppresses
  19. L21:2816 weakly suppresses
  20. L16:2579 weakly suppresses
  21. L20:2665 weakly excites
  22. L18:2399 weakly excites
  23. L25:981 weakly suppresses
  24. L11:553 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 -1.37 and -1.201 values between -1.20 and -1.021 values between -1.02 and -0.852 values between -0.85 and -0.671 values between -0.67 and -0.5040 values between -0.50 and -0.32352 values between -0.32 and -0.152077 values between -0.15 and 0.031216 values between 0.03 and 0.20181 values between 0.20 and 0.3818 values between 0.38 and 0.554 values between 0.55 and 0.732 values between 0.73 and 0.901 values between 0.90 and 1.081 values between 1.08 and 1.253 values between 1.25 and 1.431 values between 1.60 and 1.781 values between 1.95 and 2.131 values between 2.13 and 2.301 values between 2.48 and 2.651 values between 2.65 and 2.831 values between 3.00 and 3.181 values between 5.45 and 5.63offnegativepositiveoften

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. yw
  2. 我希望
  3. grap
  4. newY
  5. ACHER
  6. ไหว
  7. wished
  8. /world
  9. brig
  10. pneum

Suppresses

  1. asename
  2. 不确定性
  3. istics
  4. ijken
  5. Keyword
  6. ằng
  7. "⏎⏎⏎
  8. 存量
  9. 异议
  10. ilogue