Neuron Explorer

Qwen3-0.6B

end of document boundary

A neuron in layer 15 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 is weakly active at the marker where one web page ends and an unrelated one begins. It is faint compared with the strongest neurons, and steering did not reveal a clear effect.

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

    & Johnson knew “about the hazards associated with talcum powder, and it basically conspired to hide that fact from the public and from the medical community.”⏎What<|endoftext|>Aleja wakes up very early in the morning, as

  2. #2 peak

    a religious or educational one.⏎D<|endoftext|>Administration of Justice 104:⏎Unit Objectives: At the completion of

  3. #3 peak

    you. You will likely need to make daily visits to your provider until you are stable.⏎Treatment usually includes:⏎- Sedative drugs to help ease withdrawal symptoms⏎<|endoftext|>THE extermination of European Jewry has come to epitomise absolute evil in

  4. #4 peak

    developed techniques for masking them out of the picture or the data.⏎Some of these are applicable to digicam night photography. Rob Galbraith's⏎site<|endoftext|>World’s Biggest Beaver Dam Photo and Pictures⏎Here’s the latest update about the

  5. #5 peak

    Particles, Wiley, 1987.⏎See also: Antineutrino⏎See also: Nuclear Reactor as the Antineutrino Source<|endoftext|>Greek government-debt crisis⏎|Greek debt crisis|⏎Greek government debt crisis articles

  6. #6 peak

    to the same species. The main difference when compare to other eggs from the same period is their irregular shape, similar to that of chicken eggs.⏎Post a Comment<|endoftext|>Generally, people take coffee as energy booster also it is presumed that coffee can beat

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: it's a more practical approach to the problem, but I don't know if it's the best. It's not the same as the problem. For example, the problem might be a bit complex, but it's a bit more difficult

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

    . They litter the Faculty of Information’s Critical Making Lab, a tiny neon-red leg perched on a book shelf, and larger, white, adult-sized calf lying beside a beat-up MacBook Pro.⏎Ginger Coons, a PhD student who

  2. #2 peak

    and 20th dynasties, and were soundly defeated in a battle in the western delta.⏎After two years of peace, another, more dangerous coalition, the Sea People, a conglomeration of migrating peoples from Asia Minor

  3. #3 peak

    drop down showing the bag with 50 animals and then, the bag with 100 animals. The same can be done for handmade coats. You might

  4. #4 peak

    picturesque beach famous for its white sand and beautiful water. The beach is a popular snorkeling spot. Honeymoon Bay, featuring white sand and crystal clear water, is a quieter beach compared to Trunk Bay.⏎Hawknest Beach,

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. L12:1803 suppresses
  2. L4:1146 excites
  3. L14:2670 suppresses
  4. L13:2967 weakly suppresses
  5. L14:666 weakly excites
  6. L14:86 weakly excites
  7. L11:851 weakly excites
  8. L9:645 weakly excites
  9. L12:490 weakly suppresses
  10. L12:107 weakly excites
  11. L13:73 weakly excites
  12. L11:214 weakly excites
  13. L14:52 weakly suppresses
  14. L13:592 weakly suppresses
  15. L14:55 weakly suppresses
  16. L14:530 weakly suppresses
  17. L11:2422 weakly suppresses
  18. L14:2524 weakly suppresses
  19. L13:2258 weakly excites
  20. L14:1490 weakly suppresses
  21. L14:1033 weakly suppresses
  22. L4:1085 weakly excites
  23. L14:1729 weakly suppresses
  24. L12:723 weakly excites

Feeds later layers

  1. L17:404 strongly excites
  2. L18:154 excites
  3. L17:1489 excites
  4. L17:770 suppresses
  5. L16:2641 suppresses
  6. L19:497 excites
  7. L16:1975 excites
  8. L19:228 excites
  9. L16:862 excites
  10. L18:2816 suppresses
  11. L19:458 excites
  12. L16:1636 excites
  13. L17:962 excites
  14. L16:962 excites
  15. L16:613 excites
  16. L16:1575 excites
  17. L17:90 suppresses
  18. L16:647 excites
  19. L16:617 excites
  20. L18:421 excites
  21. L20:168 excites
  22. L19:358 excites
  23. L16:510 suppresses
  24. L18:1266 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 -2.58 and -2.461 values between -2.46 and -2.351 values between -2.12 and -2.011 values between -2.01 and -1.892 values between -1.89 and -1.781 values between -1.78 and -1.672 values between -1.67 and -1.553 values between -1.55 and -1.445 values between -1.44 and -1.3212 values between -1.32 and -1.2110 values between -1.21 and -1.1019 values between -1.10 and -0.9828 values between -0.98 and -0.8735 values between -0.87 and -0.7549 values between -0.75 and -0.6477 values between -0.64 and -0.52112 values between -0.52 and -0.41164 values between -0.41 and -0.30300 values between -0.30 and -0.18565 values between -0.18 and -0.071433 values between -0.07 and 0.05550 values between 0.05 and 0.16287 values between 0.16 and 0.28123 values between 0.28 and 0.3959 values between 0.39 and 0.5039 values between 0.50 and 0.6214 values between 0.62 and 0.732 values between 0.73 and 0.855 values between 0.85 and 0.962 values between 0.96 and 1.072 values between 1.07 and 1.191 values between 1.19 and 1.301 values between 1.42 and 1.531 values between 1.64 and 1.761 values between 1.87 and 1.99offnegativepositiveoften

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. 茫
  2. __$
  3. 奇迹
  4. --- ⏎
  5. marvel
  6. bindValue
  7. miracles
  8. 又
  9. napraw
  10. Also

Suppresses

  1. SEE
  2. .getLocal
  3. ULER
  4. QRSTUV
  5. andbox
  6. SES
  7. estic
  8. STER
  9. eload
  10. etime