Neuron Explorer

Qwen3-0.6B

unclear: fragments inside longer words

A neuron in layer 23 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 is hard to pin down. It fires on pieces of longer words or names, such as the middle of 'Polk', 'Polaris' or 'volcano', and on a few Latvian words. Turning it up or down gave repetitive or unrelated text, so it doesn't show a clear job.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    100 it gets 1st grade and is very good.⏎If the sheep gets<|endoftext|>|Republic of Latvia⏎|Anthem: Dievs, svētī Latviju!⏎God Bless Latvia!⏎and largest city⏎|Eth

  2. #2 peak

    landers, whose cloth-making industry relied on England for wool. Early in the 14th century, Flemish artisans rose up in a series of bloody revolts against the aristocratic cloth dealers who had long monopolized power. The Count

  3. #3 peak

    , the other three districts have swung between control of both parties in recent years.⏎And at the legislative level, the focus will be on suburban districts like those in Polk County, where Democrats hope booming population will lead to more favorable districts. Population

  4. #4 peak

    drained by Little Antietam Creek. The major portions of the caves are developed along strike joints, with bedding exerting an influence on passage shape and speleothem occurrence. The general dip is 25º to the east away from

  5. #5 peak

    with metal alloys for aircraft. He had the results of seven years' labour in a portfolio with him and rather than let the research fall into German hands, (Poland having been invaded by Germany), he has hidden it.<|endoftext|>BURKE TE

  6. #6 peak

    confer, like the ability to develop uncanny expertise in an area of interest. This year's class includes specialists on supervolcanoes and medieval weaponry.⏎''Look at Jack,'' Justin pointed out. ''

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: [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] [] []

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

    products. In the spring of 1934, the Franks r…⏎E-pasta adrese, uz kuru nosūtīt darba saiti:⏎Saite uz darbu:<|endoftext|>Branch of biology that

  2. #2 peak

    Old Ind. grīva- “neck,” grīvā- “the back part of the neck

  3. #3 peak

    it gets 1st grade and is very good.⏎If the sheep gets<|endoftext|>|Republic of Latvia⏎|Anthem: Dievs, svētī Latviju!⏎God Bless Latvia!⏎and largest city⏎|Ethnic groups (

  4. #4 peak

    (70 million). Other Major ones include Wu (90 Million) Min (70 million) in total 13 “sub languages” (ISO 639-3) While the Chinese character would be in

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. L21:1138 strongly excites
  2. L19:1404 suppresses
  3. L21:1519 excites
  4. L10:1921 weakly suppresses
  5. L4:1261 weakly suppresses
  6. L22:1105 weakly excites
  7. L0:989 weakly suppresses
  8. L19:1944 weakly excites
  9. L12:2740 weakly excites
  10. L1:382 weakly excites
  11. L11:1594 weakly excites
  12. L9:1990 weakly excites
  13. L1:2536 weakly suppresses
  14. L8:1174 weakly excites
  15. L20:1798 weakly suppresses
  16. L12:101 weakly excites
  17. L15:524 weakly excites
  18. L20:2166 weakly excites
  19. L8:7 weakly excites
  20. L1:136 weakly excites
  21. L1:3061 weakly excites
  22. L2:24 weakly suppresses
  23. L21:792 weakly excites
  24. L10:1083 weakly excites

Feeds later layers

  1. L25:427 weakly suppresses
  2. L26:408 weakly suppresses
  3. L24:534 weakly suppresses
  4. L27:715 weakly suppresses
  5. L25:2415 weakly excites
  6. L24:1995 weakly suppresses
  7. L24:1695 weakly excites
  8. L26:1944 weakly excites
  9. L25:962 weakly excites
  10. L26:1982 weakly excites
  11. L25:1029 weakly excites
  12. L26:1082 weakly excites
  13. L26:1288 weakly suppresses
  14. L25:1742 weakly suppresses
  15. L24:1279 weakly excites
  16. L24:44 weakly excites
  17. L24:1931 weakly excites
  18. L24:466 weakly suppresses
  19. L25:2423 weakly excites
  20. L26:2676 weakly suppresses
  21. L27:1253 weakly excites
  22. L26:112 weakly suppresses
  23. L25:475 weakly suppresses
  24. L24:667 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 -5.25 and -4.881 values between -4.52 and -4.161 values between -4.16 and -3.802 values between -2.71 and -2.353 values between -2.35 and -1.982 values between -1.98 and -1.621 values between -1.62 and -1.2618 values between -1.26 and -0.9055 values between -0.90 and -0.53408 values between -0.53 and -0.172714 values between -0.17 and 0.19596 values between 0.19 and 0.5684 values between 0.56 and 0.9211 values between 0.92 and 1.283 values between 1.28 and 1.641 values between 1.64 and 2.012 values between 2.37 and 2.731 values between 2.73 and 3.091 values between 3.46 and 3.821 values between 4.18 and 4.541 values between 5.27 and 5.631 values between 8.90 and 9.26offnegativepositiveoften

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. longing
  2. lycer
  3. ăng
  4. hare
  5. pherd
  6. agus
  7. lington
  8. 娘家
  9. veillance
  10. мес

Suppresses

  1. []
  2. []
  3. ([]
  4. [][]
  5. ([]
  6. [][]
  7. []{
  8. []⏎
  9. [] ⏎
  10. icolor