Neuron Explorer

Qwen3-0.6B

nouns for who or what a text is about

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 is unclear and loosely tied to descriptions of a subject. It fires on words like 'files', 'kids', 'audience' or 'may' in sentences about who or what a text is describing, such as 'All of these files will adhere' or 'Kids in this age group are dealing'. Turning it up or down gave repetitive or unrelated text.

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

    between the bottom-left square and the rightmost coloured square.⏎To evaluate your solution, the judges will run your program against several different input files. All of these files will adhere to the following bounds:⏎As<|endoftext|>Broccoli is a cruciferous

  2. #2 peak

    and begin to code.⏎Your program should read from the file . The file will describe a single post-cubist squarist painting.⏎Your program should

  3. #3 peak

    for this audience still tend to be based in the real world, and the chapters rely on cliffhangers to keep readers engaged. Kids in this age group are dealing with social situations, fitting in, and learning that there are consequences to their choices

  4. #4 peak

    appearance, and correct.” – Mike Markel⏎Technical writing is used in many technical and professional fields. Its goal is to objectively convey to specific audiences, who may not be subject matter experts, in an accessible manner and in easy to understand language

  5. #5 peak

    depression does not need to be addressed. Experts agree that depression, regardless of its cause, is something that should be addressed and treated.⏎Teens may also be dealing with bullying at school, or even just “harmless” teasing. They

  6. #6 peak

    application in laboratories. However, there are several factors that influence the results, such as different types of pH meters and electrodes, the varying characteristics of a sample to be measured, and the chemical properties of a solution. METTLER TOLEDO

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 can't be the only thing in the mind in the mind, it's also very important in, the mind is the thing, in the mind, the problem, in the problem, the matter, in the matter, the matter,

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

    pixel. Thus, an 8-bit pixel would have a maximum value of (2^8) - 1 = 255. Let the maximum value for pixels in image I be MAX.⏎Express the PSNR in decibels

  2. #2 peak

    91-1995⏎The materials in this archive are copyright © The Manitoba Library Association. Reproduction for personal use is permitted only if this copyright notice is maintained. Any other reproduction is prohibited without permission Copyright information for reviewers⏎Young

  3. #3 peak

    this message, Please click this link to reload this page.(Do not use your browser's "Refresh" button). Please email us if you're running the latest version of your browser and you still see this message.⏎Table of contents⏎Bluetooth is

  4. #4 peak

    A good general rule of thumb is, if they’re old enough to read, they’re old enough to pack their own bags. Assuming your little ones are old enough, the most effective

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. L11:1900 suppresses
  2. L7:61 excites
  3. L12:386 excites
  4. L11:1296 suppresses
  5. L12:1986 excites
  6. L8:584 excites
  7. L12:1938 excites
  8. L12:163 weakly suppresses
  9. L11:2224 weakly suppresses
  10. L7:611 weakly excites
  11. L11:741 weakly suppresses
  12. L7:1413 weakly excites
  13. L7:1454 weakly excites
  14. L10:1363 weakly excites
  15. L7:1657 weakly excites
  16. L6:1812 weakly suppresses
  17. L7:1381 weakly excites
  18. L5:1973 weakly excites
  19. L8:248 weakly excites
  20. L7:949 weakly suppresses
  21. L0:463 weakly suppresses
  22. L7:1024 weakly suppresses
  23. L8:2211 weakly suppresses
  24. L10:1597 weakly excites

Feeds later layers

  1. L15:1594 suppresses
  2. L14:1342 suppresses
  3. L17:2729 excites
  4. L14:2363 suppresses
  5. L19:1507 excites
  6. L17:2087 excites
  7. L17:1461 suppresses
  8. L17:2644 suppresses
  9. L18:2348 suppresses
  10. L17:2531 suppresses
  11. L14:3031 excites
  12. L14:2504 excites
  13. L15:1513 suppresses
  14. L14:1287 suppresses
  15. L16:1866 excites
  16. L16:2846 suppresses
  17. L14:685 excites
  18. L18:1606 excites
  19. L15:826 suppresses
  20. L21:2779 excites
  21. L20:2529 suppresses
  22. L15:2964 excites
  23. L15:2319 excites
  24. L14:922 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 -1.35 and -1.251 values between -1.05 and -0.961 values between -0.96 and -0.861 values between -0.86 and -0.763 values between -0.76 and -0.672 values between -0.67 and -0.5711 values between -0.57 and -0.4725 values between -0.47 and -0.3756 values between -0.37 and -0.28109 values between -0.28 and -0.18333 values between -0.18 and -0.081688 values between -0.08 and 0.02996 values between 0.02 and 0.11359 values between 0.11 and 0.21150 values between 0.21 and 0.3166 values between 0.31 and 0.4032 values between 0.40 and 0.5021 values between 0.50 and 0.6016 values between 0.60 and 0.7013 values between 0.70 and 0.793 values between 0.79 and 0.893 values between 0.89 and 0.994 values between 0.99 and 1.091 values between 1.09 and 1.184 values between 1.18 and 1.282 values between 1.28 and 1.382 values between 1.38 and 1.471 values between 1.47 and 1.571 values between 1.77 and 1.861 values between 1.86 and 1.961 values between 2.45 and 2.54offnegativepositiveoften

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. unknown
  2. unknown
  3. ars
  4. NER
  5. jas
  6. 为主线
  7. دير
  8. ncy
  9. nc
  10. ners

Suppresses

  1. _buffers
  2. forge
  3. émon
  4. iest
  5. ANDARD
  6. 就可以
  7. easiest
  8. ifact
  9. zf
  10. 便可