Neuron Explorer

Qwen3-0.6B

digits inside years and decades

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

It lights up on the middle digits of years and decades, like the '0' in 2001 or the '8' in the 1980s. Its strongest negative values come from digits in long ID numbers such as book codes. Turning it up or down gave text that did not relate to dates, so steering does not show its role.

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

    faculty member at Carnegie-Mellon University in Pittsburgh, Pennsylvania. He spent fifty years of his life, from 1949 through 2001, as a social scientist and professor of administration, psychology, and computer

  2. #2 peak

    , yes, US victory in the Cold War all played a part, because the US space program of the 1960s through the 1980s was more about perfecting technologies for defeating the Soviet Union than about pursuing

  3. #3 peak

    70s bought those semi-enclosed pedestal-style pay phones which, for most of us (Superman included) found lacking.⏎By the 1990s there were nearly three million pay phones in America. But now just a

  4. #4 peak

    , streams, groundwater).⏎Final Exam (May 17: 10:30-12:30)⏎The final exam will be on materials presented during the final quarter of

  5. #5 peak

    results⏎Authors: Walker, Robert J. (John), 1801-1869⏎Date: May 27, 1857⏎

  6. #6 peak

    for parents, knowing<|endoftext|>The N wagons were utilised for construction and maintenance of rail lines from the 1860's to the 1900's.⏎They were replaced by the _QN_ and _NN_ bog

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 can't be with you in a way that you are with me. But this is not the only way to have a positive and effective work for you, you can have all the other ways of working in a different way. But the

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

    /27/15⏎Isothermal Community College⏎P.O. Box 804, Spindale NC 28160⏎(828) 395-1307 fax (8

  2. #2 peak

    ||12.0 MB|⏎|Publisher: ||Apple Academic Press|⏎|Date published: || 2016|⏎|ISBN: ||9781771882903 (DRM-PDF)

  3. #3 peak

    Size: ||4.4 MB|⏎|Publisher: ||CRC Press|⏎|Date published: || 2015|⏎|ISBN: ||9781498738552 (DRM-PDF)

  4. #4 peak

    Frank”. The book is published by Red Fox, an imprint of Random House. (see www.kidsatrandomhouse.co.uk) ISBN 978-0-0

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. L11:1629 suppresses
  2. L17:2360 suppresses
  3. L17:2884 suppresses
  4. L17:2721 excites
  5. L17:926 weakly suppresses
  6. L14:2673 weakly suppresses
  7. L1:1245 weakly excites
  8. L17:1264 weakly excites
  9. L17:820 weakly suppresses
  10. L16:151 weakly excites
  11. L17:3006 weakly excites
  12. L12:697 weakly suppresses
  13. L18:900 weakly suppresses
  14. L8:2329 weakly suppresses
  15. L10:622 weakly excites
  16. L13:3039 weakly excites
  17. L3:2122 weakly suppresses
  18. L7:1058 weakly suppresses
  19. L11:1314 weakly suppresses
  20. L18:2519 weakly suppresses
  21. L18:2150 weakly excites
  22. L10:2892 weakly excites
  23. L6:1149 weakly suppresses
  24. L1:1446 weakly excites

Feeds later layers

  1. L20:672 weakly suppresses
  2. L21:906 weakly suppresses
  3. L20:1415 weakly suppresses slot before the second number in a range
  4. L22:823 weakly suppresses
  5. L20:1884 weakly suppresses
  6. L20:2482 weakly excites
  7. L20:1998 weakly excites
  8. L21:1746 weakly excites
  9. L24:1666 weakly suppresses
  10. L21:1198 weakly suppresses
  11. L22:610 weakly excites
  12. L22:252 weakly suppresses
  13. L20:1473 weakly suppresses
  14. L20:1709 weakly suppresses
  15. L20:924 weakly excites
  16. L22:151 weakly suppresses
  17. L25:512 weakly suppresses
  18. L21:488 weakly excites
  19. L24:467 weakly suppresses
  20. L23:2142 weakly excites
  21. L21:2784 weakly excites
  22. L23:1955 weakly excites
  23. L20:2576 weakly excites
  24. L23:799 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 -31.77 and -30.691 values between -23.19 and -22.121 values between -22.12 and -21.051 values between -14.62 and -13.554 values between -13.55 and -12.481 values between -12.48 and -11.411 values between -11.41 and -10.341 values between -10.34 and -9.262 values between -8.19 and -7.123 values between -7.12 and -6.055 values between -6.05 and -4.984 values between -4.98 and -3.914 values between -3.91 and -2.8411 values between -2.84 and -1.7640 values between -1.76 and -0.693600 values between -0.69 and 0.38196 values between 0.38 and 1.4514 values between 1.45 and 2.523 values between 2.52 and 3.594 values between 3.59 and 4.676 values between 4.67 and 5.742 values between 5.74 and 6.812 values between 6.81 and 7.881 values between 10.02 and 11.09offnegativepositiveoften

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. деся
  2. ply
  3. 几步
  4. sign
  5. committed
  6. ancer
  7. ittance
  8. 簌
  9. 玙
  10. ularity

Suppresses

  1. iminal
  2. 视听
  3. reur
  4. inate
  5. ligt
  6. 府
  7. theories
  8. imus
  9. representations
  10. ல