Neuron Explorer

Qwen3-0.6B

line break after a section heading

A neuron in layer 3 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 wakes up on the line break that comes right after a short section title in a web article, such as a heading like 'Redundancy and Backup' just before the body text starts. Turning it up didn't produce anything clearly related, so its exact role is hard to pin down.

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 more. These activities not only promote critical thinking and problem-solving skills but also foster a love for scientific inquiry.⏎Cultivate Creativity with Arts and Crafts⏎Artistic expression plays a vital role in a child’s development. Encourage your

  2. #2 peak

    it wavier and sluggish. This results in prolonged weather patterns, whether they be heatwaves or cold spells.⏎Case Study: Arctic Influence on North American Winters⏎In a recent issue of National Geographic, Arctic air is increasingly reaching southern

  3. #3 peak

    design digital circuits and systems at a higher level of abstraction, making FPGA design more efficient and less error-prone.⏎Differences Between Microcontrollers and FPGAs⏎Microcontrollers are programmable, which means that developers can write software to control the

  4. #4 peak

    entire foot, including your heels, and not just with your toes. This will prevent strain on your knee joints.⏎Burn Calories, but at a Slower Rate⏎You will burn calories using a rowing machine, but not as many as when

  5. #5 peak

    freely practice their faith without prejudice.⏎California law provides protections against discrimination based on an individual’s sexual orientation, including LGBTQ+ individuals.⏎Pregnancy and Family Status⏎Pregn<|endoftext|>Seeing Moses walk out of sight in the vaporous heights of Sinai

  6. #6 peak

    one child coercing another to follow suit. Often the behavior being pressured is questionable or taboo, such as smoking cigarettes or drinking alcohol.⏎Peer Pressure and the Brain⏎Recent studies find that peer pressure influences the way our brains behave, which leads to

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: a year ago, or a month ago, or a day ago. So I have been thinking about 20 years ago. I have been thinking about 20 years ago, 20 days ago, or 20 days

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

    Advisory Awareness among a Representative Sample of Wisconsin Adults, 2017–2019" (2022

  2. #2 peak

    any other parasitican superfamily and, to a much greater extent<|endoftext|>Is dog saliva cleaner than humans?⏎Is My Dog’s Mouth Cleaner Than a Human Mouth? The general answer to this question is no, a dog’s mouth is not technically

  3. #3 peak

    . But the code base or project is not necessarily used for the production version of the application.⏎What is a mockup logo?⏎What Is A Logo Mockup? A logo mockup is a template you can use to present your logo designs in

  4. #4 peak

    atic encephalopathy), vomiting blood or passing out altered blood in the stools. Of course, there is the dreaded complication of liver cancer.⏎Generally, cirrhosis appears in at least 20% of patients within 20 years

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. L2:1706 excites
  2. L2:156 suppresses
  3. L2:189 suppresses
  4. L0:2919 weakly suppresses
  5. L1:1441 weakly suppresses
  6. L0:667 weakly suppresses
  7. L2:373 weakly excites
  8. L2:1413 weakly excites
  9. L0:788 weakly excites
  10. L2:34 weakly excites
  11. L2:2279 weakly excites
  12. L1:2385 weakly suppresses
  13. L1:300 weakly suppresses
  14. L1:168 weakly suppresses
  15. L2:411 weakly suppresses
  16. L1:175 weakly excites
  17. L1:1097 weakly suppresses
  18. L1:2676 weakly suppresses
  19. L2:1132 weakly suppresses
  20. L2:351 weakly suppresses
  21. L1:637 weakly suppresses
  22. L0:337 weakly suppresses
  23. L1:1666 weakly excites
  24. L1:2791 weakly suppresses

Feeds later layers

  1. L5:1078 excites
  2. L4:402 excites
  3. L9:257 excites
  4. L6:451 excites line break before a subheading
  5. L6:937 excites
  6. L5:2956 excites
  7. L8:1195 weakly excites
  8. L4:43 weakly excites
  9. L15:2829 weakly excites
  10. L4:260 weakly excites
  11. L8:666 weakly excites
  12. L19:1685 weakly excites
  13. L4:1149 weakly suppresses
  14. L21:524 weakly excites
  15. L5:530 weakly excites
  16. L14:784 weakly excites
  17. L6:1013 weakly excites
  18. L15:3048 weakly excites
  19. L16:2905 weakly excites
  20. L15:975 weakly excites
  21. L5:343 weakly excites
  22. L20:2342 weakly excites
  23. L17:2137 weakly suppresses
  24. L18:983 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 -1.73 and -1.591 values between -1.45 and -1.321 values between -1.32 and -1.183 values between -1.04 and -0.901 values between -0.90 and -0.765 values between -0.76 and -0.6224 values between -0.62 and -0.4857 values between -0.48 and -0.3489 values between -0.34 and -0.20492 values between -0.20 and -0.062285 values between -0.06 and 0.08772 values between 0.08 and 0.22150 values between 0.22 and 0.3617 values between 0.36 and 0.503 values between 0.50 and 0.642 values between 0.64 and 0.781 values between 1.20 and 1.341 values between 1.34 and 1.481 values between 2.03 and 2.171 values between 2.31 and 2.451 values between 3.71 and 3.85offnegativepositiveoften

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. ificial
  2. 这个时代
  3. 绰
  4. ighted
  5. lessness
  6. 这篇文章
  7. gerekt
  8. 这个词
  9. /bit
  10. [block

Suppresses

  1. .setEnabled
  2. izens
  3. 帮他
  4. ('.')
  5. }
  6. ElementsByTagName
  7. allowed
  8. .allowed
  9. ()
  10. res