Neuron Explorer

Qwen3-0.6B

end of document token

A neuron in layer 19 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 marker the model uses to show that one web page or article has ended and a different one begins. It is strongest right where a text stops. Turning it up or down just made the model's writing collapse into repetition or nonsense.

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

    and to support supplement programs aimed at boosting the quality of education in urban settings.⏎Under Chief Justice Robert Wilentz (1979-199<|endoftext|>Tobacco Industry Influence on the U.S. Military⏎Location(s): United

  2. #2 peak

    to be debate in the party about the way forward, but its leaders have shown few signs that they are willing to give up, or even share, power.<|endoftext|>The Federal

  3. #3 peak

    by so many news sources.⏎Michelle Kaffenberger is a former research assistant at The Heritage Foundation and is currently a graduate student in Economic Development at Vanderbilt University.<|endoftext|>As a gastrointestinal surgeon and critical care specialist, John Alverdy has treated many

  4. #4 peak

    status||Published - 17 Oct 2008|⏎- age-related changes⏎- equid skulls⏎- horse use⏎- work-related changes<|endoftext|>Make your Arduino play the French Can Can using a buzzer. No pliers

  5. #5 peak

    melanoma.⏎This fact sheet from the National Cancer Institute explains the subject thoroughly. It also lists a number of the clinical trials currently underway on experimental therapeutic vaccines.<|endoftext|>What Are PK Boosters?⏎Just like with anything, gardening and⏎PK Booster

  6. #6 peak

    and Emerging Technologies Flagship competition launched by the European Commission, the other one being the “Graphene” project led by Swedish theoretical physicist Jari Kinaret.<|endoftext|>

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 a 4444-4555-5555-5555-5555-5555-5555-5555-5555

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

    civil transport aircraft designed by a UK team is leading a pack of three concepts undergoing preliminary studies for the three-year, €7 million ($9 million) European hypersonic research project, Long Term Advanced Propulsion Concepts and Technologies (LAPCAT).⏎

  2. #2 peak

    2006, 2nd ed. ) by Patriia C.⏎John Winthrops City on a hill EssayJohn winthrops city on a hill John Winthrops 1630 speech not only contributed an enduring metaphor

  3. #3 peak

    Eating disorders most commonly affect women and occur in adolescence and young adulthood, though they can occur in anyone, of any gender, at any age.”⏎Eating disorders often occur alongside other mental health disorders such as obsessive compulsive disorder, mood disorders

  4. #4 peak

    heavy-ion collisions at the Relativistic Heavy Ion Collider (RHIC), a facility that is dedicated to studying nuclear physics. These heavy strange baryons, containing at least one strange quark, still cannot be observed directly,

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. L16:562 suppresses
  2. L17:212 excites
  3. L5:46 suppresses
  4. L17:91 suppresses
  5. L17:495 excites
  6. L2:135 excites
  7. L15:219 suppresses
  8. L3:59 suppresses
  9. L17:2843 weakly suppresses
  10. L18:257 weakly suppresses
  11. L15:159 weakly excites
  12. L15:2311 weakly excites
  13. L17:183 weakly suppresses
  14. L18:260 weakly suppresses
  15. L17:102 weakly excites
  16. L14:397 weakly suppresses
  17. L3:115 weakly excites
  18. L8:2577 weakly suppresses
  19. L17:1527 weakly excites
  20. L6:7 weakly excites
  21. L17:339 weakly excites
  22. L5:57 weakly excites
  23. L15:2073 weakly suppresses
  24. L16:2508 weakly excites

Feeds later layers

  1. L27:727 weakly excites
  2. L20:628 weakly excites
  3. L21:106 weakly suppresses
  4. L21:1227 weakly suppresses
  5. L27:897 weakly excites
  6. L27:2298 weakly excites
  7. L20:681 weakly excites
  8. L22:213 weakly suppresses
  9. L27:1487 weakly excites
  10. L20:897 weakly excites
  11. L21:348 weakly excites mid-sentence line breaks
  12. L26:460 weakly suppresses
  13. L26:1518 weakly excites
  14. L27:1568 weakly excites
  15. L27:181 weakly excites
  16. L27:1118 weakly excites
  17. L27:2899 weakly excites
  18. L20:252 weakly suppresses
  19. L27:2038 weakly suppresses
  20. L27:2352 weakly excites
  21. L27:1767 weakly excites
  22. L26:380 weakly excites
  23. L20:294 weakly excites
  24. L27:2756 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.

2 values between -5.23 and -4.566 values between -3.89 and -3.2218 values between -3.22 and -2.5545 values between -2.55 and -1.88107 values between -1.88 and -1.22257 values between -1.22 and -0.55550 values between -0.55 and 0.12701 values between 0.12 and 0.79670 values between 0.79 and 1.46551 values between 1.46 and 2.13357 values between 2.13 and 2.80264 values between 2.80 and 3.47157 values between 3.47 and 4.13105 values between 4.13 and 4.8054 values between 4.80 and 5.4720 values between 5.47 and 6.1418 values between 6.14 and 6.8110 values between 6.81 and 7.486 values between 7.48 and 8.152 values between 8.15 and 8.812 values between 8.81 and 9.481 values between 10.15 and 10.821 values between 10.82 and 11.491 values between 12.16 and 12.821 values between 16.84 and 17.501 values between 18.84 and 19.511 values between 20.85 and 21.52offnegativepositiveoften

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. ...⏎⏎⏎⏎
  2. 词条
  3. nâ
  4. —⏎⏎
  5. .…⏎⏎
  6. �能
  7. â
  8. ��
  9. ries
  10. […]⏎⏎

Suppresses

  1. gaz
  2. .scalajs
  3. urdy
  4. Mc
  5. Rule
  6. ⏎
  7. BASEPATH
  8. ownt
  9. ট
  10. <<