Neuron Explorer

Qwen3-0.6B

end-of-text token after document breaks

A neuron in layer 7 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 most strongly on the hidden 'end of document' marker that separates one web page from the next. When it was turned up, the text drifted into odd, unrelated topics with no clear 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

    959,<|endoftext|>Everyone's heard about an apple a day. As one of the most popular fruits

  2. #2 peak

    criteria.<|endoftext|>The haunting image of the fallen soldier, aged just 18, is one

  3. #3 peak

    :<|endoftext|>Copyright © University of Cambridge. All rights reserved.⏎'Aim High' printed from

  4. #4 peak

    :<|endoftext|>The hamstring muscles – the biceps femoris, semitendinosus,

  5. #5 peak

    the support leg; <|endoftext|>MAHAD SATYAGRAHA AND BABASAHEB AMBEDK

  6. #6 peak

    , and other vehicles.<|endoftext|>Most people remember the adage that having a hot breakfast will set the tone for

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: - The function g() is used to represent the operation of taking the function $ f(x) $ and applying it to get the result $ f(x) $. So, the question is: when should I use the function $ f

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

    behavior. Here's a class called CarParts that declares a static inner class called Wheel. If we consider the real-world, we can find many objects

  2. #2 peak

    re toothed tiger. This adrenaline rush causes the memory to be imprinted into the amygdala, which is part of the limbic system. The amygdala holds the emotional significance of the event, including the intensity of emotion. The

  3. #3 peak

    oding – you’ve got to manage the energy you bring into a room and have a plan. With a plan, your brain is moving out of the reactive amygdala

  4. #4 peak

    counsel is grounds for automatic reversal of the conviction. Other mistakes are judged by the Harmless Error Rule. The Confrontation Clause of the Sixth Amendment gives the defense the right to cross-examine witnesses who testify about lineups, Showups,

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. L5:1234 excites
  2. L6:1803 excites
  3. L5:1650 weakly suppresses
  4. L3:2635 weakly suppresses
  5. L3:1912 weakly suppresses
  6. L6:8 weakly suppresses
  7. L3:2023 weakly suppresses
  8. L6:12 weakly excites
  9. L3:533 weakly excites
  10. L2:1657 weakly excites
  11. L6:1821 weakly suppresses
  12. L3:1101 weakly suppresses
  13. L3:551 weakly excites
  14. L5:2191 weakly excites
  15. L5:1574 weakly suppresses
  16. L1:997 weakly excites
  17. L5:102 weakly excites
  18. L0:1954 weakly excites
  19. L5:1353 weakly suppresses
  20. L0:2242 weakly excites
  21. L5:50 weakly excites
  22. L6:587 weakly suppresses
  23. L4:1882 weakly excites
  24. L6:1355 weakly suppresses

Feeds later layers

  1. L11:1159 excites
  2. L16:2172 weakly excites
  3. L9:2376 weakly suppresses
  4. L10:2352 weakly suppresses
  5. L9:1115 weakly excites
  6. L8:1853 weakly suppresses
  7. L18:2186 weakly suppresses
  8. L13:1295 weakly excites
  9. L13:786 weakly suppresses
  10. L8:1603 weakly suppresses
  11. L10:1140 weakly excites
  12. L10:2284 weakly excites
  13. L8:1096 weakly suppresses
  14. L12:2109 weakly suppresses
  15. L25:2292 weakly suppresses
  16. L9:1862 weakly excites
  17. L21:3058 weakly suppresses
  18. L11:2517 weakly suppresses
  19. L24:1800 weakly suppresses
  20. L13:2629 weakly excites
  21. L24:1022 weakly suppresses
  22. L9:2830 weakly suppresses
  23. L9:838 weakly excites
  24. L10:1321 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 -0.74 and -0.633 values between -0.52 and -0.4226 values between -0.42 and -0.31131 values between -0.31 and -0.20494 values between -0.20 and -0.091259 values between -0.09 and 0.021224 values between 0.02 and 0.13539 values between 0.13 and 0.23179 values between 0.23 and 0.3440 values between 0.34 and 0.459 values between 0.45 and 0.561 values between 0.56 and 0.671 values between 1.32 and 1.431 values between 3.49 and 3.59offnegativepositiveoften

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. Preferred
  2. Preferred
  3. preferred
  4. Armenian
  5. preferred
  6. tlement
  7. prefer
  8. ayrıl
  9. orderly
  10. ardown

Suppresses

  1. اع
  2. ity
  3. ceptor
  4. 乎
  5. ceptors
  6. 革
  7. ibu
  8. icus
  9. نو
  10. لاق