Neuron Explorer

Qwen3-0.6B

'for' before a span of time

A neuron in layer 22 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 the word 'for' when it comes before a length of time, as in 'known for years' or 'for decades'. Turned up, the model drifted into lists of numbers and the word 'many', which fits a counting theme.

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

    and challenges for these individuals truly is a spectrum that can range from a completely nonverbal kindergartner, to a professor of linguistics.⏎We’ve known for more than 20 years that the lion’s share of the risk for developing

  2. #2 peak

    kids busy. Look around you and spot something ANYTHING that either you don’t see often or haven’t seen in awhile (not a person) and get excited while smiling and saying, “OM

  3. #3 peak

    佘孟兰 Sheh Meng-lan); Mark F. Watson)<|endoftext|>By Cary Groner⏎3D printing has been waiting to disrupt the orthotics market for years. Only time will tell if the pandemic’s pause afforded clinicians the opportunity to

  4. #4 peak

    England, in 1813 and made a career for himself as a renowned architect, engineer and surveyor. After living in Dublin for several years in the early 1830s, Lanyon moved

  5. #5 peak

    Global Environment at Harvard’s School of Public Health.⏎As global temperatures rise and extreme heat waves become more common, so does heat illness. But while doctors have known for decades about the adverse effects of some medications on how the body regulates temperature, little

  6. #6 peak

    only hinted at by the book.⏎I have admired Dick’s work for years, but this is his best. He delves more thoroughly into many themes

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: 333 many many many 33 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

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

    -8 rigid rays. The long second dorsal fin has 21-24 branched, soft rays. The anal fin has one spiny ray and 24-26 soft rays. The caudal fin is truncated.⏎

  2. #2 peak

    Medical-Surgical Nursing at a Glance is ideal for nursing students looking for a clear and visual summary of care of the adult patient.⏎Results 1-5 of 41⏎The nurse practitioner is skilled at

  3. #3 peak

    color-coded congressional districts showing the average annual change in employment from 2011 to 2020. Each district within a state is given one of four distinct shades of reddish brown, with the darkest color indicating the most job

  4. #4 peak

    forests with little herbaceous growth, but with a ground cover thick with ferns, mosses, and lichens. The female is believed to give birth to a single young. The species comes down seasonally to the gullies,

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. L21:1647 strongly excites
  2. L19:2492 strongly excites "one of" before a group
  3. L18:2183 strongly suppresses
  4. L21:2090 strongly suppresses
  5. L20:1404 excites
  6. L17:2655 excites
  7. L18:1578 excites
  8. L20:1800 suppresses
  9. L20:2136 excites
  10. L14:2494 suppresses
  11. L19:2056 suppresses
  12. L20:1788 suppresses
  13. L21:822 suppresses
  14. L19:2841 suppresses
  15. L19:2503 suppresses
  16. L19:2151 excites
  17. L20:1219 suppresses
  18. L20:1787 suppresses
  19. L21:1102 excites 'one or more' before a count noun
  20. L16:2111 excites
  21. L20:2830 excites
  22. L19:2727 excites
  23. L20:2316 suppresses
  24. L19:2842 suppresses

Feeds later layers

  1. L24:826 strongly suppresses
  2. L26:1083 strongly suppresses
  3. L27:810 strongly suppresses
  4. L23:2230 suppresses
  5. L24:587 suppresses
  6. L25:572 suppresses
  7. L26:1784 suppresses
  8. L27:696 suppresses
  9. L24:2948 suppresses
  10. L25:2915 suppresses
  11. L25:2902 suppresses
  12. L24:722 suppresses 'about' or 'approximately' before a number
  13. L23:808 suppresses
  14. L25:1046 excites
  15. L23:888 suppresses
  16. L23:538 suppresses
  17. L23:1404 suppresses
  18. L24:1826 suppresses
  19. L25:2527 suppresses
  20. L26:2537 suppresses
  21. L23:1178 suppresses
  22. L27:1226 suppresses
  23. L24:680 suppresses
  24. L24:1848 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.

2 values between -11.59 and -10.802 values between -10.80 and -10.001 values between -9.20 and -8.411 values between -8.41 and -7.614 values between -7.61 and -6.827 values between -6.82 and -6.0211 values between -6.02 and -5.2216 values between -5.22 and -4.4320 values between -4.43 and -3.6330 values between -3.63 and -2.8341 values between -2.83 and -2.0477 values between -2.04 and -1.24175 values between -1.24 and -0.441594 values between -0.44 and 0.351852 values between 0.35 and 1.1545 values between 1.15 and 1.956 values between 1.95 and 2.746 values between 2.74 and 3.544 values between 3.54 and 4.341 values between 4.34 and 5.134 values between 5.13 and 5.932 values between 5.93 and 6.733 values between 6.73 and 7.521 values between 9.12 and 9.911 values between 13.10 and 13.891 values between 15.49 and 16.281 values between 19.47 and 20.27offnegativepositiveoften

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. หาร
  2. 号线
  3. tera
  4. مناسب
  5. .Helper
  6. zym
  7. hop
  8. [file
  9. лик
  10. heels

Suppresses

  1. several
  2. many
  3. 若干
  4. three
  5. 几个
  6. 多个
  7. many
  8. three
  9. two
  10. dozens