Neuron Explorer

Qwen3-0.6B

unclear: title or list start after a break

A neuron in layer 13 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the first word of a new item or title, usually right after a break in the text, like the word after a book list or a heading. The pattern is fuzzy and the steering results showed nothing clear.

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

    a beam of white rays is dispersed by a prism which colour will be refracted to a larger extent?Answer: Violet.⏎This what would happen if humans disappeared essay carries with it immense complications owing to the Suns what would happen if humans disappeared essay

  2. #2 peak

    Conference held in July, 1972, laid down the ‘National Guide lines’⏎that were to govern the ceiling<|endoftext|>Daddy Long Legs and The Inchworm Book Release⏎It wasn’t too long ago that arts education in schools

  3. #3 peak

    1983; she was the first woman executed in the United States<|endoftext|>AVTOR: JASON BLAKE⏎Who’s who? What’s This? and What Does “…” include?⏎Unclear pronouns, clipped lists, and too

  4. #4 peak

    A Memoir⏎- The Woman I Wanted to Be. Diane Von Furstenberg⏎- Dressed: A Century of Hollywood Costume Design⏎- Secrets of Stylists: An Insider's Guide

  5. #5 peak

    Resurrection fern is an intriguing species of plant known by the scientific name Pleopeltis pol

  6. #6 peak

    , giving<|endoftext|>Everyone can help to commemorate the centennial of women’s suffrage in Indiana. These resources can help you get started.⏎Preserving Women’s Legacy Grants⏎Three Indiana communities, Angola, Michigan City, and Peru, were awarded grants

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 am going to go to the store, and I can have a snack for a snack. I can have a snack for a snack for a snack for a snack for a snack for a snack for a snack for a snack for a snack

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

    be caught; levers are not capable of systematic exploits. Vote flipping, common on electronic machines, is also not possible on a lever machine. Tampering to a lever machine can be detected because the working parts of the lever machine are visible,

  2. #2 peak

    (2) A type of valve, used on some aerostats, which opens at a predetermined volume or hull dimension. maneuvering valve-A manually operated valve fitted to the envelope, ballonet, or gas cell of an aerostat for

  3. #3 peak

    .⏎Sex and gender identity or expression are protected grounds under the Act.⏎Combining the first two parts of the definition

  4. #4 peak

    the easiest question is often the toughest… Anyway, I will share my understanding of class with you all… What is a class in Java? Local variables − Variables defined inside methods, constructors or blocks are called local variables. Classes define the behavior for

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. L12:3063 suppresses
  2. L11:1290 excites
  3. L11:2385 suppresses
  4. L8:917 excites
  5. L8:781 suppresses
  6. L8:2106 excites
  7. L11:1124 suppresses
  8. L11:514 excites
  9. L10:1274 excites
  10. L10:638 excites
  11. L8:230 excites
  12. L7:2570 excites
  13. L9:544 suppresses
  14. L11:1229 weakly suppresses
  15. L4:91 weakly excites
  16. L4:1861 weakly suppresses
  17. L9:1143 weakly suppresses
  18. L6:393 weakly suppresses
  19. L7:1237 weakly suppresses
  20. L11:190 weakly excites
  21. L10:892 weakly excites
  22. L2:89 weakly suppresses
  23. L9:669 weakly suppresses
  24. L10:2026 weakly excites

Feeds later layers

  1. L14:1227 excites
  2. L14:137 excites
  3. L15:2942 excites
  4. L15:762 suppresses
  5. L15:345 suppresses
  6. L14:488 excites
  7. L14:591 excites
  8. L23:980 suppresses
  9. L14:459 weakly excites
  10. L15:1395 weakly excites
  11. L15:1551 weakly excites
  12. L16:1365 weakly suppresses
  13. L14:1998 weakly suppresses
  14. L14:37 weakly excites
  15. L15:1180 weakly suppresses
  16. L14:804 weakly excites
  17. L16:605 weakly excites
  18. L14:514 weakly excites
  19. L15:1594 weakly suppresses
  20. L14:925 weakly suppresses
  21. L17:723 weakly excites
  22. L16:2083 weakly suppresses comma after 'in this article'
  23. L15:366 weakly excites
  24. L15:1614 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 -1.67 and -1.541 values between -1.54 and -1.415 values between -1.28 and -1.146 values between -1.14 and -1.0111 values between -1.01 and -0.8814 values between -0.88 and -0.7435 values between -0.74 and -0.6171 values between -0.61 and -0.48145 values between -0.48 and -0.35249 values between -0.35 and -0.21605 values between -0.21 and -0.081548 values between -0.08 and 0.05806 values between 0.05 and 0.19271 values between 0.19 and 0.3280 values between 0.32 and 0.4521 values between 0.45 and 0.5910 values between 0.59 and 0.726 values between 0.72 and 0.856 values between 0.85 and 0.983 values between 0.98 and 1.124 values between 1.12 and 1.253 values between 1.25 and 1.382 values between 1.38 and 1.521 values between 1.52 and 1.652 values between 1.65 and 1.781 values between 1.78 and 1.911 values between 3.51 and 3.64offnegativepositiveoften

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. 什么意思
  2. 是一家
  3. сть
  4. 是一款
  5. 額
  6. trình
  7. اف
  8. ↩
  9. 是一位
  10. 年起

Suppresses

  1. erten
  2. nerRadius
  3. IMPLIED
  4. arming
  5. enaries
  6. atedRoute
  7. .gstatic
  8. osen
  9. erguson
  10. ang