Neuron Explorer

Qwen3-0.6B

surname split after first letter D

A neuron in layer 24 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 letter D when it starts a surname that the tokenizer has split in two, as in 'Jason Dworkin' or 'Gwynne Dyer'. Turning it up pushed the model toward unrelated, rambling text, so the steering does not show a clear effect.

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

    be available for study for future generations, it is possible the person who names the asteroid will grow up to study the regolith we return to Earth,” said Jason Dworkin, OSIRIS-REx project scientist at NASA’s Godd

  2. #2 peak

    after a baby’s birth, a period that has been dubbed the “magic moment.” The new study is the first to test that assumption, said author Christina Gibson-Davis. “It turns out the ‘magic moment’ lasts longer than conventional wisdom

  3. #3 peak

    found that most children born out of wedlock don’t remain so: 64 percent of children born out of wedlock see their moms get married, Gibson-Davis said. Many of those marriages don’t last, however. Nearly half of

  4. #4 peak

    subject. Hughes argues that Shakespeare's concern with Lucrece's chastity is reflective of a larger metaphysical concern with the sacredness of the soul. Carolyn D. Williams (1993) examines Lucrece's rape, as

  5. #5 peak

    heading into deep space.⏎“The mission was essentially over by<|endoftext|>Human nature and the nature of war⏎Twenty years ago, Canadian military historian and journalist Gwynne Dyer fascinated the public in 45 countries with his award-winning television series

  6. #6 peak

    going to be another productive year in physical education class!<|endoftext|>Spring 2010 ACCESS #36⏎Just Road Pricing by Lisa Schweitzer and Brian D. Taylor⏎Economists have long advocated road pricing as an efficient way to

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: what if I can actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actually actual? I have to check to see if this actual of actual of actual of actual of actual of actual of actual

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

    قاله پژوهشی⏎1 دانشیار معماری، دانشکده معماری و شهرسازی، دانشگاه علم و صنعت ایران، تهران

  2. #2 peak

    شگر دوره تخصصی دکتری معماری منظر، دانشکده معماری و شهرسازی، دانشگاه علم و صنعت ایران، تهران

  3. #3 peak

    engrin⏎There are straws in the wind, a growing discontent with the French.Here Be Dragons⏎Trends of 'discontent'⏎View usage over:⏎'discontent' in Other Languages⏎Translation of discontent from the Collins English to Portuguese

  4. #4 peak

    "Here we show that Epo increases the motivation to exercise," said Max Gassmann, D.V.M., a researcher involved in the work from the Institute of Veterinary Physiology,

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. L23:400 excites
  2. L22:1511 suppresses
  3. L2:1125 weakly excites
  4. L23:89 weakly suppresses
  5. L19:251 weakly excites
  6. L18:2364 weakly suppresses
  7. L20:259 weakly suppresses
  8. L23:86 weakly suppresses
  9. L18:569 weakly excites
  10. L4:2965 weakly excites
  11. L15:1893 weakly excites
  12. L14:1884 weakly suppresses
  13. L21:237 weakly suppresses
  14. L3:952 weakly excites
  15. L6:2812 weakly excites
  16. L22:2665 weakly suppresses physical measurements and dimensions
  17. L10:1673 weakly excites
  18. L10:1774 weakly suppresses
  19. L22:793 weakly suppresses
  20. L19:841 weakly excites
  21. L14:2928 weakly suppresses
  22. L20:2381 weakly suppresses
  23. L6:240 weakly suppresses
  24. L0:371 weakly excites

Feeds later layers

  1. L27:1955 excites
  2. L25:1285 suppresses
  3. L25:1817 weakly suppresses
  4. L25:572 weakly suppresses
  5. L25:1814 weakly suppresses
  6. L27:874 weakly suppresses
  7. L26:1396 weakly suppresses
  8. L25:1142 weakly suppresses weight loss and metabolism claims
  9. L27:2852 weakly suppresses
  10. L27:1796 weakly suppresses
  11. L25:649 weakly suppresses
  12. L25:2488 weakly excites
  13. L26:1040 weakly suppresses
  14. L27:1522 weakly suppresses
  15. L25:1418 weakly excites
  16. L27:1976 weakly suppresses
  17. L27:686 weakly excites
  18. L26:2020 weakly excites
  19. L26:2095 weakly suppresses
  20. L25:1703 weakly suppresses
  21. L27:555 weakly excites
  22. L25:2118 weakly suppresses
  23. L25:408 weakly excites
  24. L26:2014 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 -24.73 and -23.881 values between -14.51 and -13.651 values between -11.95 and -11.101 values between -11.10 and -10.252 values between -10.25 and -9.391 values between -9.39 and -8.543 values between -8.54 and -7.691 values between -7.69 and -6.841 values between -5.98 and -5.132 values between -5.13 and -4.285 values between -4.28 and -3.433 values between -3.43 and -2.573 values between -2.57 and -1.7217 values between -1.72 and -0.871327 values between -0.87 and -0.022493 values between -0.02 and 0.8436 values between 0.84 and 1.695 values between 1.69 and 2.541 values between 2.54 and 3.391 values between 4.25 and 5.101 values between 5.10 and 5.951 values between 7.66 and 8.511 values between 8.51 and 9.36offnegativepositiveoften

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. мы
  2. úa
  3. dın
  4. positories
  5. стер
  6. 再也不
  7. UV
  8. ASON
  9. erra
  10. 不做

Suppresses

  1. anden
  2. ector
  3. toInt
  4. ern
  5. )-(
  6. entifier
  7. 对象
  8. (type
  9. .allocate
  10. 拊