Neuron Explorer

Qwen3-0.6B

comma after a clause opener

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 the comma that comes right after an introductory phrase like 'However,' or 'In winters,' and expects a new clause to follow. When it was turned up, the text drifted into rambling, unrelated sentences.

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

    If you are pondering how to make your computer virus, you’ve come towards the right place. There are many tutorials and steps on making a malware. However , prior to this, it is important to understand what the pathogen is, just

  2. #2 peak

    the young.<|endoftext|>Our bodies constitute various types of cells and perform different functions depending upon their location. All cell types have their special characteristics and functions,for example , we have the germ cells which are responsible for the formation of spermatozoa

  3. #3 peak

    see there are so many questions asked meanwhile and students are expected to answers them. Some students may face some problem while answering it , but we promise you that after reviewing our answer, you will gain the confidence to

  4. #4 peak

    malware didn’t spread on its own, this set the stage meant for other dangerous laptop viruses to follow. However , this is just a few samples of the dangers of computer infections. In fact ,

  5. #5 peak

    value 0 on entry to method another() so printing these values gives 10 and 0 respectively. 3)The original object ref v is unchanged , it

  6. #6 peak

    Extra) Exercise⏎Class 10 Science Chapter 4 Environmental managementNotes:⏎As we go through the entire Class 10 Science Chapter 4 Environmental management , we can

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: some wellmented, wellmented, wellmented, wellmented, wellmented, wellmented, wellmented, wellmented, wellmented, wellmented, just so, just quiet, just quiet, just quiet, just quiet, just quiet

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

    bag at one of the bucket drain holes. Let the water drain out slowly near the trunk, but not on<|endoftext|>Electric cars may be energy efficient Earth-saving futuremobiles, but their rise could also have a strange unintended consequence: killing off the

  2. #2 peak

    as the academic community. There is little precedent for this kind of open operating model.⏎According to project manager Pasi Kankaanpää a massive need foropen access to imaging facilities can be seen across the entire science community and imaging is among

  3. #3 peak

    term birth, as well as density of Superfund Sites in the United States. As part of NIEHS's Superfund Research Program, the Puerto Rico Testsite for Exploring Contamination Threats (PROTECT) is exploring the relationships

  4. #4 peak

    So here it is! The Little Dipper! The North Star, Polaris, marks the end of its handle.⏎Written by EarthSky⏎Other Links:Print This Post<|endoftext|>One basic concept of an operating system is the process. If we

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. L20:763 suppresses
  2. L11:29 suppresses
  3. L16:248 suppresses
  4. L16:1543 excites
  5. L20:959 suppresses
  6. L2:49 excites
  7. L2:1321 excites
  8. L18:396 excites
  9. L18:546 excites
  10. L19:251 suppresses
  11. L18:576 excites
  12. L13:129 excites
  13. L5:50 suppresses
  14. L20:131 excites
  15. L7:22 excites
  16. L15:819 weakly suppresses
  17. L19:697 weakly excites
  18. L2:75 weakly excites
  19. L6:429 weakly suppresses
  20. L3:421 weakly excites
  21. L18:853 weakly excites
  22. L14:313 weakly suppresses
  23. L18:45 weakly suppresses
  24. L18:569 weakly suppresses

Feeds later layers

  1. L23:889 strongly suppresses
  2. L24:2026 strongly suppresses
  3. L25:2562 strongly suppresses
  4. L25:2286 strongly excites
  5. L25:561 suppresses
  6. L27:2491 suppresses
  7. L26:3030 suppresses
  8. L27:2769 weakly suppresses capital 'K' starting a name piece
  9. L27:170 weakly suppresses unclear: spaces in mixed-language text
  10. L23:1252 weakly excites
  11. L27:2767 weakly suppresses
  12. L23:474 weakly excites
  13. L25:956 weakly excites
  14. L23:1037 weakly excites
  15. L24:938 weakly excites
  16. L26:1142 weakly suppresses
  17. L27:343 weakly excites
  18. L23:550 weakly excites
  19. L26:810 weakly excites
  20. L23:2881 weakly excites
  21. L27:30 weakly excites
  22. L26:609 weakly excites
  23. L23:1486 weakly excites
  24. L24:2060 weakly 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.

1 values between -25.28 and -24.211 values between -21.00 and -19.931 values between -18.86 and -17.791 values between -17.79 and -16.721 values between -16.72 and -15.652 values between -15.65 and -14.572 values between -14.57 and -13.503 values between -13.50 and -12.438 values between -12.43 and -11.3610 values between -11.36 and -10.2917 values between -10.29 and -9.2230 values between -9.22 and -8.1537 values between -8.15 and -7.0850 values between -7.08 and -6.0191 values between -6.01 and -4.94127 values between -4.94 and -3.87217 values between -3.87 and -2.80344 values between -2.80 and -1.73432 values between -1.73 and -0.66504 values between -0.66 and 0.42459 values between 0.42 and 1.49457 values between 1.49 and 2.56355 values between 2.56 and 3.63249 values between 3.63 and 4.70199 values between 4.70 and 5.77121 values between 5.77 and 6.8471 values between 6.84 and 7.9145 values between 7.91 and 8.9828 values between 8.98 and 10.0516 values between 10.05 and 11.1214 values between 11.12 and 12.197 values between 12.19 and 13.262 values between 13.26 and 14.343 values between 14.34 and 15.411 values between 15.41 and 16.482 values between 16.48 and 17.55offnegativepositiveoften

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. adays
  2. etheless
  3. xiety
  4. alyze
  5. atomy
  6. phalt
  7. tlement
  8. bsites
  9. alog
  10. icularly

Suppresses

  1. auses
  2. ////////////////////////////////////////////////////////////////
  3. handled
  4. ÜR
  5. deceptive
  6. appable
  7. mercado
  8. Ngu
  9. BUM
  10. dap