Neuron Explorer

Qwen3-0.6B

commas inside short lists

A neuron in layer 16 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 reacts to commas that separate items in a short list of three or four things, such as 'motivation, confidence, arousal'. Its strongest negative cases are also commas in lists, so the pattern is not fully clear. Turning it up gave 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

    pertaining to the four active branches of the military (Army, Navy, Marines, Air Force); 2) Examining current tobacco policies and practices in

  2. #2 peak

    That's<|endoftext|>In applied sport psychology, there are four mental areas that are considered to be central for athletes to be mentally prepared for athletic success: motivation, confidence, arousal, and concentration.

  3. #3 peak

    ' are trademarks of The Financial Times Ltd.⏎We live in an era of big data – often characterised in terms of the three “v”s: volume, velocity and variety.⏎Three simple examples illustrate the phenomenon. Volume – Tesco has

  4. #4 peak

    0 annotated summer driving thermal images captured at day and night and their corresponding RGB imagery for reference. The dataset includes classification of five groups; people, dogs, cars, bicycles

  5. #5 peak

    wine, he commanded the seven eunuchs who served him – Methuan, Biztha, Harbon, Bigtha, Anagtha, Zethat and Karkas – to bring before him Queen Vashti,

  6. #6 peak

    is controlled by the ‘analogWrite’ function.⏎Do RGB LEDs need PWM?⏎an RGB LED is a combination of three LEDs in just one package: red, green and blue; there are two kinds of RGB LEDs: common cathode and

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, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the,

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

    , men were believed to grow beards for different reasons than men do today. The three main reasons prehistoric men had beards were for warmth, intimidation, and to defend their face from hits. In Egyptian times, men would dye their

  2. #2 peak

    this blog is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of qualified health providers with questions you may have

  3. #3 peak

    a small incision in your belly. This procedure is generally performed under general anesthesia in a hospital.⏎- MRI. This test uses a large magnet, radio waves, and a computer to make 2-D images of an internal

  4. #4 peak

    Me" (NBC), "Ally McBeal" (FOX), "That '70s Show" (FOX), "The Simpsons" (FOX), and "Dawson's Creek" (FOX). Taylor said these programs were

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. L13:1066 strongly suppresses
  2. L15:2500 excites 'and' closing a list of items
  3. L15:941 excites
  4. L10:667 suppresses commas inside lists of names
  5. L15:2598 suppresses
  6. L15:2999 suppresses
  7. L15:1029 excites comma in three-item lists
  8. L14:35 suppresses
  9. L15:137 suppresses
  10. L15:132 excites
  11. L14:449 excites
  12. L8:1605 excites
  13. L14:2531 excites
  14. L15:345 suppresses
  15. L13:1296 excites
  16. L14:2657 suppresses
  17. L9:77 excites commas in citations and name lists
  18. L15:274 excites
  19. L13:557 excites
  20. L14:907 excites
  21. L9:2447 suppresses
  22. L3:372 excites
  23. L12:2723 weakly suppresses
  24. L13:269 weakly suppresses

Feeds later layers

  1. L19:492 excites
  2. L17:1802 excites
  3. L17:2654 excites
  4. L21:393 suppresses unclear: relative 'that' and list items
  5. L20:1009 excites
  6. L20:882 suppresses
  7. L19:1089 suppresses
  8. L17:572 weakly excites
  9. L27:358 weakly suppresses
  10. L19:628 weakly excites
  11. L18:2921 weakly excites period ending a sentence that announces a list
  12. L26:236 weakly suppresses
  13. L17:2610 weakly excites and or comma closing a list
  14. L17:919 weakly suppresses
  15. L18:1533 weakly suppresses
  16. L17:878 weakly excites
  17. L18:2863 weakly suppresses
  18. L20:282 weakly suppresses
  19. L21:154 weakly excites
  20. L26:22 weakly excites the 'res' fragment inside words
  21. L20:2864 weakly excites
  22. L20:2429 weakly excites
  23. L24:435 weakly suppresses
  24. L23:673 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 -8.61 and -8.313 values between -5.88 and -5.581 values between -5.58 and -5.271 values between -5.27 and -4.973 values between -4.97 and -4.675 values between -4.67 and -4.369 values between -4.36 and -4.066 values between -4.06 and -3.768 values between -3.76 and -3.4510 values between -3.45 and -3.1512 values between -3.15 and -2.858 values between -2.85 and -2.548 values between -2.54 and -2.2410 values between -2.24 and -1.9419 values between -1.94 and -1.6318 values between -1.63 and -1.3337 values between -1.33 and -1.03100 values between -1.03 and -0.72230 values between -0.72 and -0.42711 values between -0.42 and -0.122080 values between -0.12 and 0.19402 values between 0.19 and 0.49110 values between 0.49 and 0.7956 values between 0.79 and 1.1026 values between 1.10 and 1.4014 values between 1.40 and 1.709 values between 1.70 and 2.013 values between 2.01 and 2.314 values between 2.31 and 2.612 values between 2.92 and 3.222 values between 3.22 and 3.52offnegativepositiveoften

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. [/
  2. lak
  3. adians
  4. AndPassword
  5. ain
  6. elsius
  7. andre
  8. pou
  9. above
  10. liness

Suppresses

  1. ially
  2. nels
  3. dict
  4. urally
  5. quoi
  6. Vers
  7. vely
  8. tps
  9. 至
  10. 箱