Neuron Explorer

Qwen3-0.6B

sentence end before a line break

A neuron in layer 24 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 at the end of a sentence or paragraph, on the full stop or comma just before the text moves to a new line. When it was turned up strongly, the model produced an endless run of blank line breaks.

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

    , delivering a smooth user experience. With the ever-growing demand for innovative apps, proficient coding skills have become invaluable for developers looking to make a mark in the industry.⏎The Challenges of Coding for Apps⏎Coding for apps presents its fair share of challenges

  2. #2 peak

    the most effective and efficient way to get things done. In this essay, we will discuss the differences between hard work and smart work and analyze the benefits of each.⏎Hard work is often associated with putting in long hours and extensive effort to achieve a

  3. #3 peak

    computer scientists got involved—at that point, things got downright ugly. As painful as it was for the computer scientists involved, the country is far better off for it.⏎The professional computing community is very familiar with the role of standards. Well-known standards

  4. #4 peak

    28, part of the project will be dedicated to the development of a exaflop supercomputer that will be used for the simulation of the brain model.⏎Brain researchers are generating 60,000 papers per year,” said

  5. #5 peak

    create a function in Arduino is vital for any programmer out there. With that said, we shall be discussing how to create a function in Arduino in today’s article.⏎In this article, you will learn how you can write functions and use them in

  6. #6 peak

    dirty and start tinkering! With a little bit of patience and dedication, you'll soon start to recognize the patterns and symbols and become an expert in no time.⏎A Brief On Kirchhoff S Laws With Circuit Diagram⏎Circuit Topology

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 fact that, perhaps, the mind is a kind of kind of resource — perhaps — rather than a kind of resource — perhaps — rather than a kind of mind — perhaps — rather than a kind of mind. Maybe the mind is a

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

    conservation area.⏎Neil burn Drive was created.⏎Canmore place was erected.<|endoftext|>ERIC Number: ED379961⏎Record Type: RIE⏎Publication Date: 19

  2. #2 peak

    fuel, chemical, and water meters.⏎Tools & Technology⏎Tools used in this occupation:⏎- Air dryers — Biomass dryers⏎- Batching plants or feeders — Reclaim feeders⏎- Belt convey

  3. #3 peak

    |⏎|Born||December 7, 1928⏎|Era||20th / 21st-century philosophy|⏎|

  4. #4 peak

    . extremes in temperatures, sunlight, tight-fitting clothing, etc.)<|endoftext|>ERIC Number: ED348648⏎Record Type: Non-Journal⏎Publication Date: 1992-Apr⏎Reference Count: N/A⏎

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. L19:187 strongly suppresses
  2. L18:2922 strongly suppresses
  3. L18:5 suppresses
  4. L20:1085 suppresses
  5. L17:829 suppresses
  6. L20:1081 excites
  7. L16:588 excites
  8. L19:441 excites
  9. L5:2007 excites
  10. L16:49 suppresses
  11. L13:10 excites
  12. L10:84 suppresses
  13. L5:203 excites
  14. L21:130 excites
  15. L15:81 suppresses line breaks inside running text
  16. L14:56 suppresses document ends and list line breaks
  17. L6:67 excites
  18. L3:70 excites
  19. L1:1279 suppresses
  20. L16:359 suppresses
  21. L5:238 excites
  22. L6:2030 excites
  23. L17:1568 suppresses
  24. L1:883 excites

Feeds later layers

  1. L27:1496 strongly excites
  2. L25:40 strongly suppresses
  3. L27:487 strongly excites
  4. L25:533 excites first letter m or oct fragments
  5. L27:1801 excites
  6. L25:91 excites
  7. L26:241 excites
  8. L26:156 suppresses
  9. L25:144 excites
  10. L25:248 suppresses
  11. L25:207 suppresses
  12. L27:256 excites
  13. L26:230 excites
  14. L27:200 suppresses
  15. L26:167 excites
  16. L26:1018 excites
  17. L26:1180 suppresses
  18. L25:297 excites
  19. L26:42 weakly suppresses
  20. L26:133 weakly suppresses
  21. L27:156 weakly suppresses
  22. L26:288 weakly suppresses
  23. L26:16 weakly excites
  24. L27:1305 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 -3.49 and -2.951 values between -2.95 and -2.403 values between -2.40 and -1.865 values between -1.86 and -1.3111 values between -1.31 and -0.77145 values between -0.77 and -0.233530 values between -0.23 and 0.32151 values between 0.32 and 0.867 values between 0.86 and 1.414 values between 1.41 and 1.953 values between 1.95 and 2.502 values between 2.50 and 3.041 values between 3.04 and 3.592 values between 3.59 and 4.132 values between 4.13 and 4.675 values between 4.67 and 5.222 values between 5.22 and 5.761 values between 5.76 and 6.314 values between 6.31 and 6.851 values between 6.85 and 7.402 values between 7.40 and 7.942 values between 7.94 and 8.481 values between 8.48 and 9.036 values between 9.03 and 9.573 values between 10.66 and 11.212 values between 11.21 and 11.751 values between 11.75 and 12.291 values between 12.29 and 12.841 values between 13.38 and 13.931 values between 13.93 and 14.472 values between 14.47 and 15.021 values between 15.56 and 16.101 values between 16.10 and 16.651 values between 16.65 and 17.192 values between 17.74 and 18.28offnegativepositiveoften

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. ⏎
  4. <|endoftext|>
  5. >⏎
  6. %%⏎
  7. __⏎
  8. ({⏎
  9. ⏎

Suppresses

  1. ?⏎⏎
  2. ?⏎⏎
  3. *****⏎⏎
  4. ?"⏎⏎
  5. ?”⏎⏎
  6. ues
  7. "?⏎⏎
  8. ******⏎⏎
  9. !.⏎⏎
  10. ?)⏎⏎