Neuron Explorer

Qwen3-0.6B

mid-word pieces like 'advent' and 'precaution'

A neuron in layer 0 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

The neuron fires on the middle pieces of certain longer words, especially 'vent' in 'Advent' and 'ca' in 'Precaution'. It is negative on pieces of medical and research terms such as 'sclerosis' and 'behavioral'. Turning it up or down changed the continuation, but the new text had no obvious link to these words.

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

    Includes information on different styles of greenhouses, solar vs. supplemental heat, figuring costs, selecting a location, glazing, environmental controls<|endoftext|>Advent Reflections: Week 2, Day 5⏎“Moses said to

  2. #2 peak

    each module.⏎Share this pageToggle animations⏎Ninjas vs Superbugs⏎Adventures in NanomedicineWatch the video⏎- When Alexander Fleming discovered penic

  3. #3 peak

    Mystery' Bird Illness Continues, Tri-State May Put Bird Feeders Back Out With Precautions⏎The agency recommends thoroughly cleaning feeders at least once a week with a

  4. #4 peak

    , please remember to carry a converter!⏎Precaution 2: Choosing a power bank⏎So far, the standard voltage of

  5. #5 peak

    do not match, it may cause the appliance to burn or be damaged!⏎Precaution 1: When traveling⏎US sockets are generally 110V

  6. #6 peak

    when you're online to keep your identity safe.⏎- Only Make Purchases On Trusted Sites⏎When deals seem too good to be true, they just may

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 have been thinking about myself, and I have been thinking about myself. It's been thinking about myself. It's been thinking about myself. It's been thinking about myself. I have been thinking about myself. It's been thinking about

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

    ogenesis by modulating host immunity through the regulation of multiple metabolic pathways.⏎The “Gut Feeling”: Breaking Down the Role of Gut Microbiome in Multiple Sclerosis published in Neurotherapeutics gives an excellent overview of the body of knowledge being explored

  2. #2 peak

    help. The sooner you receive care for your anxiety symptoms, the better. Feel free to reach out to Neuro-Behavioral Clinical Research for help and information on anxiety.⏎RECEIVE A $7

  3. #3 peak

    can be used to<|endoftext|>| Trisha Sertori |⏎JUGA, Indonesia (Reuters) – The last time Wayan Rendeh took part in a temple ceremony

  4. #4 peak

    and treatments, rehabilitation, counseling, a full-time caretaker, inability to drive<|endoftext|>UK hydrogen cars are coming - if you can fill up⏎LONDON (Reuters) - Britain's hydrogen fuel cell car fleet may hit top gear within five 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

None: this is the first layer.

Feeds later layers

  1. L4:3010 suppresses
  2. L3:936 suppresses
  3. L11:302 weakly suppresses
  4. L1:1958 weakly excites
  5. L17:2949 weakly suppresses 'for' introducing an example or reason
  6. L9:1850 weakly suppresses
  7. L2:2800 weakly suppresses
  8. L5:2032 weakly excites
  9. L6:470 weakly suppresses
  10. L22:1357 weakly excites
  11. L15:921 weakly excites
  12. L18:186 weakly suppresses
  13. L5:1124 weakly suppresses
  14. L15:477 weakly suppresses
  15. L4:2218 weakly excites
  16. L24:1532 weakly excites
  17. L7:390 weakly suppresses
  18. L18:2521 weakly suppresses
  19. L10:2970 weakly excites
  20. L19:1289 weakly suppresses
  21. L24:2594 weakly suppresses
  22. L4:365 weakly excites
  23. L2:1826 weakly excites
  24. L10:2412 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 -0.49 and -0.462 values between -0.46 and -0.441 values between -0.44 and -0.411 values between -0.36 and -0.342 values between -0.34 and -0.313 values between -0.31 and -0.292 values between -0.29 and -0.272 values between -0.27 and -0.246 values between -0.24 and -0.2211 values between -0.22 and -0.1926 values between -0.19 and -0.1747 values between -0.17 and -0.1481 values between -0.14 and -0.12151 values between -0.12 and -0.09247 values between -0.09 and -0.07365 values between -0.07 and -0.05590 values between -0.05 and -0.021149 values between -0.02 and 0.00613 values between 0.00 and 0.03271 values between 0.03 and 0.05148 values between 0.05 and 0.0883 values between 0.08 and 0.1046 values between 0.10 and 0.1327 values between 0.13 and 0.1517 values between 0.15 and 0.186 values between 0.18 and 0.204 values between 0.20 and 0.233 values between 0.25 and 0.271 values between 0.27 and 0.301 values between 0.40 and 0.421 values between 0.47 and 0.50offnegativepositiveoften

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. dojo
  2. edo
  3. UserDao
  4. uche
  5. aldi
  6. bsites
  7. udo
  8. Wrest
  9. фор
  10. ELY

Suppresses

  1. sav
  2. _endian
  3. thing
  4. rest
  5. increased
  6. thing
  7. increases
  8. �单
  9. multiple
  10. phục