Neuron Explorer

Qwen3-0.6B

colon before a subtitle

A neuron in layer 22 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 a colon that separates a title from its subtitle, like the colon in a book or paper title. When it was turned up, the model wrote a lot of 'A:' and 'I:' labels, which matches its focus on colons.

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

    .L., & REIJMERSDAL, E.A. (2012). Social connections and the persuasiveness of viral campaigns in social network sites: Persuasive intent as the underlying mechanism. Journal of Marketing, 18

  2. #2 peak

    reasoning is used as the context in which the contributors elaborate on their own perceptions.⏎Table of Contents⏎Contents: W.F. Overton, Competence and Procedures: Constraints on the Development of Logical Reasoning. B. Inhelder,

  3. #3 peak

    gencies, and International Intervention Chapter 3 Gender and the Causes and Consequences of Armed Conflict Chapter 4 Gender Mainstreaming in United Nations Peacekeeping Operations: Talking the Talk, Tripping over the Walk Part 5 Gender, Peacekeeping

  4. #4 peak

    ing, Logic, and Thought Disorders: Deductive Reasoning and Developmental Psychopathology. D.P. Keating, Structuralism, Deconstruction, Reconstruction: The Limits of Reasoning.⏎".....Of interest to serious advanced students of cognitive

  5. #5 peak

    9, the study, “Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the Global Burden of Disease Study 2013

  6. #6 peak

    of outbreak response⏎Public Health Responses to a Dengue Outbreak in a Fragile State: A Case Study of Nepal⏎This case study illustrates the principles taught in Lecture

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: how he so ...he so he so he so he so he so he so he so he so he so he so he so he so he so he so he

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

    victory of good over evil. Dussehra or Vijaya Dashami or Dasara is marked on the 10th day of the bright half (Shukla Paksha) of the month of Ashvin (Ashwayuja),

  2. #2 peak

    by Muhammad Ali Jinnah from the Muslim league, B R Ambedkar campaigning for the rights of Dalits, and Master Tara Singh from SAD (Shiromani Akali Deal). This was the first-ever articulation for the

  3. #3 peak

    Government of Nepal gave the mountain the official Nepali name. In sanskrit Sagarmatha means "mother of the universe”. The Tibetan name for Mount Everest is Chomolungma or

  4. #4 peak

    stices, equinoxes, and phases of the moon for Vedic rituals. Eighteen early astronomical texts or siddhantas, of which only the Surya-Siddhantha, written around 400 BCE, survives

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. L17:1362 excites unclear: book and report titles
  2. L20:1016 excites
  3. L12:2494 suppresses
  4. L19:2302 suppresses
  5. L15:2532 suppresses
  6. L13:2123 weakly excites
  7. L15:1066 weakly suppresses
  8. L4:30 weakly excites
  9. L18:1338 weakly excites
  10. L15:2459 weakly suppresses
  11. L10:2801 weakly excites
  12. L14:1909 weakly excites
  13. L10:270 weakly suppresses
  14. L5:2838 weakly excites
  15. L15:345 weakly excites
  16. L5:1440 weakly suppresses
  17. L13:1325 weakly excites
  18. L7:1356 weakly suppresses
  19. L17:621 weakly suppresses
  20. L10:2002 weakly excites
  21. L3:233 weakly excites
  22. L1:2083 weakly excites
  23. L16:1652 weakly suppresses
  24. L10:1661 weakly suppresses

Feeds later layers

  1. L23:236 excites
  2. L25:745 excites
  3. L26:648 suppresses
  4. L24:322 excites
  5. L26:122 excites
  6. L25:45 suppresses
  7. L27:858 excites
  8. L25:4 suppresses
  9. L25:733 weakly excites
  10. L27:1103 weakly excites
  11. L27:663 weakly excites
  12. L26:582 weakly excites
  13. L25:837 weakly excites
  14. L27:646 weakly excites
  15. L26:1229 weakly suppresses
  16. L27:960 weakly excites
  17. L24:864 weakly suppresses 'to' before an infinitive
  18. L24:823 weakly excites
  19. L27:797 weakly excites
  20. L25:440 weakly suppresses
  21. L25:1132 weakly suppresses
  22. L26:2432 weakly suppresses
  23. L25:1586 weakly excites
  24. L26:778 weakly excites capital letters inside acronyms

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 -11.09 and -9.961 values between -9.96 and -8.831 values between -8.83 and -7.711 values between -6.58 and -5.451 values between -5.45 and -4.321 values between -4.32 and -3.196 values between -3.19 and -2.06116 values between -2.06 and -0.933178 values between -0.93 and 0.20484 values between 0.20 and 1.3352 values between 1.33 and 2.4619 values between 2.46 and 3.5914 values between 3.59 and 4.724 values between 4.72 and 5.856 values between 5.85 and 6.984 values between 6.98 and 8.111 values between 8.11 and 9.244 values between 10.37 and 11.502 values between 11.50 and 12.632 values between 12.63 and 13.763 values between 13.76 and 14.894 values between 16.02 and 17.151 values between 26.19 and 27.321 values between 27.32 and 28.451 values between 32.96 and 34.09offnegativepositiveoften

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. :
  3. :a
  4. :A
  5. :X
  6. :user
  7. part
  8. :Is
  9. :end
  10. :x

Suppresses

  1. 率为
  2. .HandlerFunc
  3. Handle
  4. 严格按照
  5. 着他
  6. ⎝
  7. titleLabel
  8. Verfügung
  9. hers
  10. 情况进行