Neuron Explorer

Qwen3-0.6B

end-of-document token

A neuron in layer 14 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 marker that signals the end of a web page or document, especially when the text stops abruptly in the middle of a list or sentence. Turning it up made the model ramble in loops, so the steering does not show a 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

    conscious you can make passwords stronger by:⏎- Use both upper and lower case letters⏎- Use more special characters⏎- Make the way you combine them more complicated<|endoftext|>If unwed parents are going to get married, the best window of opportunity for

  2. #2 peak

    Particles, Wiley, 1987.⏎See also: Antineutrino⏎See also: Nuclear Reactor as the Antineutrino Source<|endoftext|>Greek government-debt crisis⏎|Greek debt crisis|⏎Greek government debt crisis articles

  3. #3 peak

    manage operations in your warehouse.⏎- Lay out a proper strategy for handling damaged products and returns from customers.⏎- Implement pest control measures to protect your products.⏎-<|endoftext|>Nucleation properties of respirable coal dust (terminal progress report).⏎Physics Department

  4. #4 peak

    . 2008. English.⏎Last Reviewed: March 2010⏎Last Updated: November 13, 2014<|endoftext|>If we wanted to create a form of feedback that encouraged students to focus less about

  5. #5 peak

    in desired direction.⏎Step back and look at pruned shrub from different angles to check final shape of shrub. It should look like a small tree.<|endoftext|>Pace has always been a leader in environmental studies, so why not take advantage

  6. #6 peak

    01.0yyyyyyy.yyyyyyyy => 172.17.0.0/17⏎- We need a network with <|endoftext|>|Classification||Electronic musical instrument, synthesizer|⏎|Inventor(s)||

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: I have been, or been in a position, or the time, or the time, or the time, or the time, or the time, or the time, or the time, or the time, or the time, or the

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

    few years ago when I was giving tours at the Utah Museum of Natural History in Salt Lake City. A fossil hunter near Kemmerer, Wyoming, had been quite surprised to find a small mammal while digging through layers of fossilized fish in

  2. #2 peak

    <|endoftext|>Re export or Re-exportation is a form of international trade in which a country exports goods previously imported from another country.⏎Okay, does this definition sound way too elaborate? Then for ease of understanding and usage, consider the following example to comprehend

  3. #3 peak

    brake pedal pressure denotes how much braking power is manual versus how much electrical energy is recaptured. This means that the harder the pedal is pressed, then the more effort is used to stop car

  4. #4 peak

    on my nightstand for several weeks. In my mind, it was going to be another book about visual learners and strategies for using images to increase engagement. I WAS WRONG. This book is different. It is not about visual learning;

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. L13:506 excites
  2. L11:1350 excites
  3. L13:181 excites
  4. L13:2522 suppresses
  5. L10:2001 suppresses
  6. L12:279 excites
  7. L12:2623 suppresses
  8. L13:1147 excites
  9. L12:1524 suppresses
  10. L12:1604 suppresses
  11. L13:2919 suppresses words leading into a planned action
  12. L11:592 excites
  13. L13:2802 excites
  14. L12:1982 excites
  15. L7:104 suppresses
  16. L10:2602 suppresses
  17. L12:884 suppresses
  18. L10:1447 suppresses
  19. L12:515 suppresses
  20. L13:491 suppresses
  21. L12:282 suppresses
  22. L12:296 suppresses
  23. L12:949 excites
  24. L8:141 weakly suppresses

Feeds later layers

  1. L16:1287 excites
  2. L16:1514 excites
  3. L17:1448 excites
  4. L16:2762 excites
  5. L16:2550 suppresses
  6. L19:1938 excites
  7. L17:381 suppresses
  8. L15:1133 suppresses
  9. L18:1537 excites
  10. L18:2235 suppresses
  11. L15:2295 suppresses
  12. L16:1207 suppresses
  13. L17:2145 suppresses
  14. L16:403 suppresses
  15. L17:3024 excites
  16. L15:1513 excites
  17. L16:1567 excites
  18. L16:3023 suppresses
  19. L15:2627 suppresses
  20. L15:1642 suppresses
  21. L16:702 suppresses
  22. L15:2284 suppresses
  23. L15:2208 excites
  24. L16:2167 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.

4 values between -1.24 and -1.112 values between -1.11 and -0.9810 values between -0.98 and -0.8620 values between -0.86 and -0.7329 values between -0.73 and -0.6160 values between -0.61 and -0.4897 values between -0.48 and -0.35184 values between -0.35 and -0.23342 values between -0.23 and -0.10975 values between -0.10 and 0.02783 values between 0.02 and 0.15462 values between 0.15 and 0.28283 values between 0.28 and 0.40199 values between 0.40 and 0.53137 values between 0.53 and 0.6578 values between 0.65 and 0.7871 values between 0.78 and 0.9047 values between 0.90 and 1.0333 values between 1.03 and 1.1618 values between 1.16 and 1.2819 values between 1.28 and 1.4112 values between 1.41 and 1.5312 values between 1.53 and 1.666 values between 1.66 and 1.797 values between 1.79 and 1.913 values between 1.91 and 2.041 values between 2.04 and 2.165 values between 2.16 and 2.294 values between 2.29 and 2.411 values between 2.41 and 2.542 values between 2.54 and 2.671 values between 2.79 and 2.921 values between 3.67 and 3.80offnegativepositiveoften

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. RESULT
  2. result
  3. resulting
  4. EndInit
  5. iert
  6. ier
  7. offspring
  8. 结果
  9. lift
  10. 結果

Suppresses

  1. itu
  2. pré
  3. ksen
  4. ufs
  5. contractual
  6. ún
  7. fores
  8. posição
  9. SingleNode
  10. Roose