Neuron Explorer

Qwen3-0.6B

word-start 'acc' fragments

A neuron in layer 19 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the opening 'acc' of words like accident, accurate and accompanied. Turning it up made the model produce strange repeated text, such as a long string of zeros, which does not reveal much about the neuron.

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

    tisation.⏎ALBERT HOFMANN AND LSD⏎Swiss research chemist for Sandoz⏎First created LSD in 1938⏎Accidentally absorbed LSD through fingers in 1943⏎Took deliberate higher

  2. #2 peak

    that can lead to drowning.⏎It also leads to an attitude of accountability and consequences. Blaming bodies of water and warm weather would lead to an attitude of “accidents happen”.⏎Icy roads are no more a cause of crashes than deep

  3. #3 peak

    . In his own words, Captain Laudonniere states: “Having reconnoitered the river, I landed to talk […]⏎AN ARTIST WHO ACCOMPANIED THE FRENCH EXPEDITION TO FLORIDA UNDER LA

  4. #4 peak

    For the most part, nonetheless, a writer retains the single right to the title, characters, story as well as storyline of the work and also is complimentary to accredit any of those elements to others. In publishing, an author gives the author

  5. #5 peak

    Terzan 5 and others like it were once small galaxies in their own right.⏎It seems the Milky Way's massive bulge (and others like it) accrete over time as dwarf galaxies erode and are eaten up by passers-by

  6. #6 peak

    recent decision by the accountancy firm Ernst and Young who decided to drop degree classification thresholds for graduate<|endoftext|>Synchronize and Simplify Barometric Compensation Across Entire Site⏎Accurate Barometric Compensation⏎Leveloggers measure absolute pressure (water pressure + atmospheric

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 user is a young, perhaps not very young, 6-8-year-old, who is a child of both of their young, both of their 5-6ths, and is not quite too young to be on the fr

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

    . What is the size of Source and Destination IP address in IP header?⏎A. 4 bits⏎B. 8 bits⏎C. 16 bits⏎D. 32 bits⏎6-7

  2. #2 peak

    is doubled?⏎A) 2 F⏎B) F / 2⏎C) F / 4⏎D) 4 F⏎E) F /

  3. #3 peak

    . This website works best with modern browsers such as the latest versions of Chrome, Firefox, Safari, and Edge. If you continue with this browser, you

  4. #4 peak

    :<|endoftext|>Copyright © University of Cambridge. All rights reserved.⏎'Aim High' printed from http://nrich.math

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. L2:1661 strongly excites
  2. L5:149 strongly suppresses
  3. L18:88 strongly excites
  4. L7:342 strongly excites
  5. L17:241 strongly suppresses
  6. L10:269 strongly excites joining word in a list
  7. L1:11 strongly suppresses
  8. L17:437 strongly excites
  9. L0:1010 strongly excites
  10. L18:440 strongly excites
  11. L17:1216 suppresses
  12. L16:374 suppresses
  13. L1:50 excites
  14. L1:340 excites
  15. L0:2910 excites
  16. L1:53 suppresses
  17. L0:1996 excites
  18. L18:288 excites
  19. L7:674 suppresses
  20. L15:114 excites
  21. L4:969 excites the letter C in vitamin C
  22. L14:643 suppresses
  23. L6:944 excites
  24. L0:2939 suppresses

Feeds later layers

  1. L22:607 strongly excites
  2. L22:1155 strongly excites
  3. L20:665 strongly excites
  4. L21:2906 strongly excites
  5. L21:590 excites
  6. L22:2898 excites
  7. L21:320 excites
  8. L26:2677 excites
  9. L20:314 excites
  10. L23:619 excites
  11. L27:2987 excites letter c inside a split word
  12. L20:613 excites
  13. L20:582 excites
  14. L26:2459 excites letter k starting a unit or fragment
  15. L22:2025 excites
  16. L25:2319 excites
  17. L22:1086 suppresses
  18. L26:1729 excites
  19. L27:2597 excites
  20. L26:1536 excites
  21. L27:2867 excites
  22. L24:2013 excites
  23. L22:2170 excites
  24. L26:2504 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.

2 values between -10.88 and -9.912 values between -7.96 and -6.993 values between -6.01 and -5.041 values between -5.04 and -4.064 values between -4.06 and -3.094 values between -3.09 and -2.11288 values between -2.11 and -1.142560 values between -1.14 and -0.16403 values between -0.16 and 0.81182 values between 0.81 and 1.79127 values between 1.79 and 2.7691 values between 2.76 and 3.7375 values between 3.73 and 4.7149 values between 4.71 and 5.6823 values between 5.68 and 6.6628 values between 6.66 and 7.6321 values between 7.63 and 8.6112 values between 8.61 and 9.5811 values between 9.58 and 10.555 values between 10.55 and 11.533 values between 11.53 and 12.503 values between 12.50 and 13.481 values between 13.48 and 14.453 values between 14.45 and 15.431 values between 15.43 and 16.403 values between 16.40 and 17.381 values between 17.38 and 18.351 values between 24.20 and 25.171 values between 27.12 and 28.09offnegativepositiveoften

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. illion
  2. YLES
  3. rypt
  4. 슭
  5. rust
  6. ipient
  7. istani
  8. opper
  9. пас
  10. .Headers

Suppresses

  1. unre
  2. enticate
  3. 国有资产
  4. rec
  5. kim
  6. ocr
  7. 出来了
  8. ⏎
  9. 沱
  10. %p