Neuron Explorer

Qwen3-0.6B

'hands' after 'wash your'

A neuron in layer 9 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 responds to the word 'hands' in advice about washing, such as 'Wash your hands often'. The strongest negative hits are on words like 'life' and 'speech'. Steering did not produce a clear theme.

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

    unethical activities as more acceptable and reasonable than they would if they had not washed themselves.⏎AdvertisementTo reach the surprising conclusion, 22 people who had washed their hands, and 22 who had not, were made to watch a disgusting scene

  2. #2 peak

    and vegetable wash is only effective if you employ all necessary safety precautions. If you touch a clean apple with dirty hands, your efforts are ruined. Always wash your hands before touching clean produce. Peel and cut your fruits and vegetables after they have been

  3. #3 peak

    a week, even if it's just a short walk or online class.⏎Get as much sleep as you can, including naps if possible.⏎Wash your hands often and avoid touching your eyes to avoid catching a virus, especially if your baby

  4. #4 peak

    can peel them yourself.⏎• Drink beverages that are bottled and have a seal that is unbroken.⏎• Wash your hands often with soap and water or alcohol-based hand sanitizer.⏎• Bugs (including mosquitoes

  5. #5 peak

    and ensure prompt and appropriate action is taken to protect the health of all Albertans.⏎You can protect yourself from all coronaviruses by:⏎- Cleaning your hands frequently using either soap and warm water or alcohol-based hand sanitizers,⏎- Cover

  6. #6 peak

    Have you checked for updates?⏎After you've run the<|endoftext|>Beginning May 9, every person older than 12 years old will be required to wear a face covering when in public anywhere in Boulder County where social distancing of 6 feet cannot

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 more I have seen, the more I have felt, the more I have learned, the more I have grown up, and the more I have experienced, the more I have been able to connect with others, and the more I have

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

    and continues to demonstrate repeatedly that the theory has no merit in reality.⏎Laboratory experiments and probabilistic calculations have definitely made it clear that the amino acids from which life arises cannot have been formed by chance. the cell, which supposedly emerged by chance

  2. #2 peak

    – the number of notes per second and the frequency – it’s just as complicated as human speech. They’re just not saying much.”⏎All the same, they’re getting their

  3. #3 peak

    00 International Units of Vitamin A and 1.2 mg of Vitamin C.<|endoftext|>A twig on the tree of life⏎Over three billion years ago, life started as a seed that grew and evolved, becoming a great tree with many branches

  4. #4 peak

    could hold clues to the origin of the solar system and organic molecules that may have seeded life on Earth. NASA also is planning a crewed mission to an asteroid by

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. L7:2877 suppresses
  2. L7:2532 excites
  3. L5:403 excites
  4. L8:1102 excites
  5. L0:463 weakly suppresses
  6. L4:1120 weakly suppresses
  7. L8:1324 weakly suppresses
  8. L6:1951 weakly excites
  9. L0:2774 weakly suppresses
  10. L4:2357 weakly excites
  11. L6:779 weakly excites
  12. L0:1816 weakly suppresses
  13. L0:2385 weakly suppresses
  14. L7:2931 weakly suppresses
  15. L6:2239 weakly excites
  16. L4:1160 weakly excites
  17. L8:2333 weakly suppresses
  18. L5:2449 weakly excites
  19. L5:961 weakly suppresses 'area' and 'volume' as measures
  20. L0:386 weakly excites
  21. L7:2852 weakly suppresses
  22. L8:2252 weakly suppresses
  23. L6:1566 weakly suppresses
  24. L8:1442 weakly excites

Feeds later layers

  1. L10:154 suppresses
  2. L10:1827 suppresses
  3. L15:1256 suppresses
  4. L11:966 suppresses
  5. L12:1204 weakly excites
  6. L10:2645 weakly suppresses
  7. L13:1019 weakly suppresses
  8. L11:1400 weakly excites
  9. L11:80 weakly suppresses
  10. L11:1324 weakly suppresses
  11. L12:1906 weakly excites
  12. L22:629 weakly suppresses
  13. L12:948 weakly suppresses
  14. L16:2753 weakly suppresses
  15. L14:1357 weakly suppresses
  16. L10:168 weakly suppresses
  17. L10:2543 weakly suppresses
  18. L12:630 weakly suppresses
  19. L11:60 weakly excites
  20. L13:2896 weakly suppresses
  21. L22:293 weakly suppresses
  22. L13:2704 weakly excites
  23. L11:1805 weakly suppresses
  24. L12:828 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 -3.85 and -3.701 values between -3.54 and -3.391 values between -3.24 and -3.094 values between -2.32 and -2.172 values between -2.17 and -2.022 values between -1.87 and -1.713 values between -1.71 and -1.562 values between -1.56 and -1.4110 values between -1.41 and -1.265 values between -1.26 and -1.1010 values between -1.10 and -0.9512 values between -0.95 and -0.8019 values between -0.80 and -0.6440 values between -0.64 and -0.4981 values between -0.49 and -0.34249 values between -0.34 and -0.19885 values between -0.19 and -0.031864 values between -0.03 and 0.12415 values between 0.12 and 0.27140 values between 0.27 and 0.4257 values between 0.42 and 0.5835 values between 0.58 and 0.7326 values between 0.73 and 0.8814 values between 0.88 and 1.0410 values between 1.04 and 1.199 values between 1.19 and 1.343 values between 1.34 and 1.493 values between 1.49 and 1.651 values between 1.65 and 1.802 values between 1.80 and 1.951 values between 1.95 and 2.101 values between 2.10 and 2.26offnegativepositiveoften

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. raison
  2. くり
  3. clientId
  4. ذا
  5. iciente
  6. .scheme
  7. anka
  8. Din
  9. 槎
  10. ulin

Suppresses

  1. lobal
  2. decode
  3. 为之
  4. 布尔
  5. roman
  6. AssertionError
  7. ORN
  8. 耶
  9. BLACK
  10. ,next