Neuron Explorer

Qwen3-0.6B

zeros inside long number codes

A neuron in layer 21 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 lights up on digits inside long number codes such as ISBN numbers, postal codes, journal ISSNs and DOIs, especially the zeros. It goes quiet on ordinary words, and it goes quiet on the word 'Opt' in phrases like 'opt-in'. Turning it up made the model's output drift into unrelated text, so the steering results don't show a clear meaning.

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

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

  2. #2 peak

    induced changes in gene flow patterns among Tasmanian devil populations, Biological Conservation, 165 pp. 69-78. ISSN 0006-320

  3. #3 peak

    provider:|| Postal: Katajanokanlaituri 6B, 00160 Helsinki|⏎Web page: http://www.wider.unu.edu

  4. #4 peak

    23(2), 1976 pp. 125–135. doi:10.1112/S0025579300008718.<|endoftext|>

  5. #5 peak

    Northwest, on the Touch for map. Marker is at or near this postal address: 3048 R Street Northwest, Washington DC 20007, United States of

  6. #6 peak

    from Wiktionary, Creative Commons Attribution/Share-Alike License⏎- n. An underwater explorer.⏎from WordNet 3.0 Copyright 2006 by Princeton University. All rights reserved.⏎- n. an underwater swimmer

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 you have a square and you need to go to a point in the square, how do you get to that point in the first and second and third and fourth and ... etc. and so on. So, first, I think of

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 and Prayer raise our frequency by 15 MHz⏎20 Ways to Increase Your Frequency & Optimize Your Health⏎As certain patterns of vibration are associated with good health, one

  2. #2 peak

    to increase walking, bicycling and using public transportation. The plan will engage the public and partners across the region to identify the region’s principal active transportation network.⏎Opt-In poll for active transportation⏎Metro recently conducted an Opt-In poll on active transportation

  3. #3 peak

    I had been shooting whistlers (golden-eye ducks) over decoys.”⏎“The software can also automatically detect and remove red - and golden-eye defects.”⏎“OptiML Red-Eye uses embedded image processing automatically detects and corrects red-

  4. #4 peak

    conditioner's ability to remove heat and lower temperature<|endoftext|>Last month, the National Academies of Science, Engineering, and Medicine hosted a 3-day workshop: "Optimizing care systems for people with intellectual and developmental disabilities." During this workshop,

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. L18:474 strongly suppresses
  2. L19:906 strongly excites
  3. L18:112 strongly suppresses 'over' meaning time or duration
  4. L10:1052 strongly suppresses
  5. L2:1795 strongly suppresses
  6. L5:1546 strongly suppresses
  7. L16:177 strongly excites zero digit inside dates and codes
  8. L17:213 strongly suppresses
  9. L0:44 suppresses
  10. L1:15 suppresses
  11. L0:1906 suppresses
  12. L17:106 excites
  13. L20:121 excites
  14. L17:1046 suppresses
  15. L5:138 suppresses
  16. L1:178 suppresses
  17. L4:498 suppresses
  18. L0:1241 excites
  19. L10:715 suppresses
  20. L1:1082 excites
  21. L7:2095 suppresses
  22. L7:1581 suppresses the last zero in large numbers
  23. L16:75 excites
  24. L2:657 suppresses

Feeds later layers

  1. L27:2618 suppresses
  2. L23:1773 suppresses
  3. L23:2981 suppresses
  4. L26:1712 suppresses single letter starting a split word
  5. L24:2973 suppresses
  6. L26:2206 weakly suppresses
  7. L26:2017 weakly suppresses
  8. L24:2537 weakly suppresses
  9. L25:2782 weakly suppresses
  10. L26:2888 weakly suppresses
  11. L24:2265 weakly excites
  12. L25:1060 weakly suppresses
  13. L27:3015 weakly suppresses 'F' starting a word in a split
  14. L23:1608 weakly suppresses
  15. L23:2515 weakly suppresses
  16. L23:1865 weakly suppresses
  17. L27:2187 weakly suppresses
  18. L23:2766 weakly suppresses
  19. L27:3054 weakly suppresses
  20. L26:1196 weakly suppresses
  21. L22:1892 weakly suppresses the digit 6 in years 19xx
  22. L25:2342 weakly suppresses
  23. L23:1366 weakly excites
  24. L24:2801 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 -114.00 and -110.961 values between -110.96 and -107.921 values between -98.80 and -95.771 values between -68.41 and -65.371 values between -44.10 and -41.062 values between -41.06 and -38.021 values between -38.02 and -34.984 values between -34.98 and -31.942 values between -31.94 and -28.904 values between -28.90 and -25.863 values between -25.86 and -22.831 values between -22.83 and -19.794 values between -19.79 and -16.752 values between -16.75 and -13.713 values between -13.71 and -10.677 values between -10.67 and -7.6313 values between -7.63 and -4.5933 values between -4.59 and -1.553289 values between -1.55 and 1.49534 values between 1.49 and 4.531 values between 4.53 and 7.57offnegativepositiveoften

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. 掌声
  2. ftware
  3. orks
  4. ости
  5. ��
  6. ость
  7. blank
  8. ulates
  9. dge
  10. olist

Suppresses

  1. entimes
  2. course
  3. lahoma
  4. posing
  5. portun
  6. imum
  7. erves
  8. ngoing
  9. 凿
  10. course