Neuron Explorer

Qwen3-0.6B

word-start 'l' after a space

A neuron in layer 26 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 when a word begins with the letter l after a space, as in 'a lopsided trade' or 'a lurch to the right'. The steering outputs were unrelated to this, so the pattern is only clear from the examples.

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

    agitated to keep the United States out of the war against Hitler, the Irish-Americans who bankrolled the IRA’s “armed struggle”, or the Cuban-Americans who lobbied to keep the US shackled to a pointless embargo of the island,

  2. #2 peak

    routes that account for the bulk of trade between the nations. A lack of quality roads around ports, insufficient warehouses, high tariffs and visa restrictions have contributed to a lopsided and lackluster trade relationship. To help ease trade barriers, negot

  3. #3 peak

    did not tangibly improve. This and the widespread anti-immigration sentiment stoked, but by no means instigated by Trump, fits this same pattern of a lurch to the right in the public mood.<|endoftext|>Food Safety: Serving⏎You

  4. #4 peak

    for action.⏎One hundred extras charged up the hill; another hundred came storming down the same hill to do mock battle. In another location Roman centurions lashed and shouted at two hundred slaves who labored to move a huge stone monument

  5. #5 peak

    the rest is used to<|endoftext|>The turn of the twentieth century is often referred to as the nadir of race relations in the United States. Jim Crow laws, lynchings, and other forms of racial violence plagued the southern region of the country

  6. #6 peak

    advanced against Egypt by land and by sea. Ramses’ land army checked the enemy’s advance in southernmost Palestine, and the hostile ships were trapped after being lured into the waterways of the delta. Egypt averted conquest by the nor

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: instead of using the abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract abstract

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

    (Trachypithecus geei),⏎- Asiatic Water Buffalo (Bubalus bubalis),⏎- Guar (Bos gaurus), Rhino (Rhinoceros unicorns),⏎- Pigmy Hog (Sus salvanius).⏎

  2. #2 peak

    this method to find out exactly what goes wrong in the nerve cells in movement disorders such as ataxias," said Prof. Dr. Stefan Herlitze (RUB Department for Biology and Biotechnology).⏎The researchers report is in the Journal of

  3. #3 peak

    ’s nocturnes, Bach’s preludes and fugues, and the works of Clementi, Mozart and Beethoven

  4. #4 peak

    in 1657. In the 18th century the Lorraine family kept collecting and the Grand Duke Peter Leopold of Habsburg-Lorraine founded in Torrigiani Palace a veritable Museum of Physics. The adjoining

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. L17:134 suppresses
  2. L5:1294 excites
  3. L18:420 excites
  4. L0:662 suppresses
  5. L21:252 excites
  6. L18:198 suppresses
  7. L2:1518 suppresses
  8. L10:273 suppresses
  9. L7:2731 excites
  10. L22:961 suppresses
  11. L20:257 weakly suppresses
  12. L2:146 weakly excites unclear: 'lesson' and 'lique' word pieces
  13. L18:331 weakly suppresses
  14. L16:83 weakly suppresses
  15. L5:1712 weakly excites
  16. L19:2084 weakly excites
  17. L19:593 weakly excites
  18. L7:421 weakly suppresses
  19. L23:1760 weakly suppresses
  20. L2:1724 weakly suppresses
  21. L8:599 weakly suppresses
  22. L20:1230 weakly excites
  23. L18:365 weakly suppresses
  24. L21:2568 weakly suppresses

Feeds later layers

  1. L27:2675 weakly excites
  2. L27:1583 weakly excites
  3. L27:1323 weakly suppresses
  4. L27:2313 weakly excites unclear: 'g' starts of words and hyphenated 'g' pieces
  5. L27:2069 weakly excites
  6. L27:1792 weakly excites
  7. L27:503 weakly suppresses
  8. L27:2010 weakly excites
  9. L27:767 weakly excites
  10. L27:2640 weakly suppresses
  11. L27:636 weakly excites
  12. L27:1482 weakly suppresses
  13. L27:1805 weakly suppresses
  14. L27:1301 weakly suppresses
  15. L27:3070 weakly excites
  16. L27:2763 weakly excites
  17. L27:2917 weakly suppresses
  18. L27:1608 weakly excites
  19. L27:665 weakly suppresses
  20. L27:2760 weakly excites 'the' after 'between' in age ranges
  21. L27:2682 weakly suppresses
  22. L27:1385 weakly excites
  23. L27:147 weakly excites
  24. L27:1784 weakly suppresses

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 -55.56 and -53.771 values between -26.89 and -25.101 values between -23.30 and -21.511 values between -16.13 and -14.341 values between -14.34 and -12.552 values between -12.55 and -10.765 values between -8.97 and -7.173 values between -7.17 and -5.385 values between -5.38 and -3.5915 values between -3.59 and -1.802601 values between -1.80 and -0.011214 values between -0.01 and 1.7919 values between 1.79 and 3.5815 values between 3.58 and 5.379 values between 5.37 and 7.168 values between 7.16 and 8.961 values between 8.96 and 10.754 values between 10.75 and 12.542 values between 14.33 and 16.13offnegativepositiveoften

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. yard
  2. ipo
  3. ây
  4. 객
  5. unittest
  6. 练
  7. #import
  8. 缦
  9. 简易
  10. *)[

Suppresses

  1. บาท
  2. vaz
  3. CVE
  4. 璋
  5. جماع
  6. etti
  7. anged
  8. crown
  9. title
  10. itre