Neuron Explorer

Qwen3-0.6B

end of document boundary

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 responds to the invisible marker the training data uses to show that one web page has ended and another has begun. It fires on the spot right before a new, unrelated text starts. Turning it up did not produce a clear theme in the model's writing.

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

    Reducer coupling nuts offer the ability to couple different rod sizes. Threaded rods are often

  2. #2 peak

    locate.<|endoftext|>Science Fair Project Encyclopedia⏎The World Factbook⏎The World Factbook is an

  3. #3 peak

    the<|endoftext|>Nature has given us all the possible amenities to keep ourselves healthy. Interestingly, our

  4. #4 peak

    the<|endoftext|>This Article explores the legacy of the U.S. environmental justice movement for the U

  5. #5 peak

    it was celebrated for a<|endoftext|>Learn something new every day⏎More Info... by email⏎A quesadilla

  6. #6 peak

    would pass away before<|endoftext|>The Badrouzi Festival or “Kajin” was held in ancient Iran

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: when you think of the natural forerites, which is the general public's natural or naturalistic, but when you think of the social forerites, which is the general public's social or socialish, but when you 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

    are several shell recycling drop-off sites across the coast including at the North Carolina Coastal Federation located at 3609 Hwy 24, Ocean, NC 28570 and at Hammocks Beach State Park located at 1

  2. #2 peak

    015 The University of North Carolina Press⏎116 South Boundary Street, Chapel Hill, NC 27514-3808⏎How to Order | Make

  3. #3 peak

    Planetarium (who's internet⏎presence is limited to⏎StarHustler at the moment) usually has set up an event⏎hotline (Phone: 305-854-2222) and features shows

  4. #4 peak

    ooteka (John Jolly) of the Cherokee, negotiated the January 14, 1836, treaty between Chief Bowl of the Cherokee and the Republic of Texas.

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. L20:431 excites
  2. L18:363 excites
  3. L16:1023 excites
  4. L3:46 suppresses
  5. L3:500 suppresses
  6. L8:407 excites
  7. L20:508 excites
  8. L16:1295 excites
  9. L20:1402 weakly suppresses
  10. L18:1203 weakly excites end of document marker
  11. L15:1165 weakly suppresses
  12. L2:1152 weakly excites
  13. L18:529 weakly excites
  14. L6:2570 weakly excites
  15. L4:2084 weakly excites
  16. L20:1829 weakly suppresses
  17. L4:1700 weakly suppresses conservation laws and invariance
  18. L14:1495 weakly suppresses
  19. L9:2169 weakly excites
  20. L20:85 weakly suppresses
  21. L20:1053 weakly suppresses
  22. L19:989 weakly excites
  23. L20:2663 weakly suppresses
  24. L9:114 weakly suppresses

Feeds later layers

  1. L25:465 weakly suppresses
  2. L26:2004 weakly suppresses
  3. L24:2919 weakly suppresses
  4. L22:1905 weakly excites
  5. L23:1539 weakly excites
  6. L27:1436 weakly excites
  7. L22:2564 weakly suppresses perpendicular and geometry terms
  8. L23:1648 weakly suppresses
  9. L23:2704 weakly suppresses
  10. L23:232 weakly excites
  11. L24:1450 weakly suppresses
  12. L24:1979 weakly suppresses
  13. L22:1790 weakly excites
  14. L25:2883 weakly suppresses
  15. L24:1511 weakly suppresses
  16. L24:1060 weakly suppresses
  17. L23:618 weakly suppresses
  18. L22:2670 weakly suppresses
  19. L23:2728 weakly suppresses second digit of years like 2020
  20. L22:499 weakly excites
  21. L23:907 weakly suppresses
  22. L25:1544 weakly suppresses
  23. L24:2146 weakly excites
  24. L24:1847 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 -11.41 and -10.861 values between -4.90 and -4.362 values between -4.36 and -3.821 values between -3.82 and -3.283 values between -2.73 and -2.191 values between -2.19 and -1.655 values between -1.65 and -1.1117 values between -1.11 and -0.571252 values between -0.57 and -0.022486 values between -0.02 and 0.52121 values between 0.52 and 1.0611 values between 1.06 and 1.603 values between 1.60 and 2.142 values between 2.14 and 2.691 values between 6.48 and 7.021 values between 9.73 and 10.27offnegativepositiveoften

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. const
  2. const
  3. →const
  4. ({...
  5. 스크
  6. cry
  7. plemented
  8. ,const
  9. interface
  10. <const

Suppresses

  1. .Connect
  2. ?⏎
  3. ije
  4. ?⏎
  5. ||⏎
  6. hormone
  7. ?
  8. 吗
  9. ?
  10. otional