Neuron Explorer

Qwen3-0.6B

"all" after a comma in a trailing phrase

A neuron in layer 18 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

This neuron lights up on the word "all" when it follows a comma and starts a tacked-on phrase, like ", all of which" or ", all without a license". It goes quiet, or negative, on words like "everyone" and "Everything" that start a new clause. Turning it up made the model's continuations wander into unrelated, rambling text, so the steering shows no 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

    New Advent and get the full contents of this website as an instant download or CD-ROM. Includes the Catholic Encyclopedia, Church Fathers, Summa, Bible and more all for only $19.99...⏎A Spanish historian; born at Cu

  2. #2 peak

    Jones, a member of the Jehovah’s Witnesses, with violating its licensing ordinance by selling books, operating as a book agent, and operating as a transient dealer, all without a license. The trial court found the Witness guilty of displaying pamphlets in

  3. #3 peak

    ’s University Belfast, United Kingdom. He received his B.Tech from CUSAT and his M.Tech and PhD from Indian Institute of Technology Madras, all in Computer Science. His current research interests include data analytics, natural language processing,

  4. #4 peak

    pressure to mental health disorders can affect the ability to regulate heat. They can alter the body’s ability to sweat, change how it regulates temperature or cause dehydration, all impairing people’s capacity to deal with rising temperatures and heat waves.⏎Doctors may

  5. #5 peak

    single integrated circuit that is designed to control specific hardware functions. It consists of a central processing unit (CPU), memory, input/output peripherals, and various interfaces, all integrated into a single chip. Microcontrollers are programmable, which means that developers

  6. #6 peak

    yourself - reflect on why. What is really pushing your buttons? Consider, are you trying to control a situation, a person’s views, or their behavior, all of which are not within your control? Is there a power struggle in the making

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: both the same, except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except except

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

    cause you to feel anxious, stressed, worried, sad, bored, lonely or frustrated.⏎It's important to remember it is OK to feel this way and that everyone reacts differently. Remember, this situation is temporary and, for most of us,

  2. #2 peak

    balances are an integral part of mental health diagnosis and management; no matter what the psychiatric disorder. Julie Foster, a nurse practitioner in Portland, Oregon notes, "Everything a person eats affects all aspects of physical and mental health and thus a dietary and

  3. #3 peak

    treasure hunters enter into highly charged territory. We had no such<|endoftext|>Everyone stands in front of a 24 x 36 sheet of newsprint

  4. #4 peak

    This is the trick you should apply. Do you want a garden contoured hill (hill), normal (flat), terraced, or variations? But remember, everything should fit your needs.⏎Afterwards, do not forget to notice the condition and

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. L11:267 excites
  2. L4:262 excites
  3. L3:2187 suppresses
  4. L13:1024 suppresses
  5. L11:583 excites 'both' after a paired subject
  6. L14:2463 excites
  7. L16:1071 suppresses "any" in questions and negatives
  8. L17:2268 excites
  9. L14:2292 excites
  10. L16:1718 suppresses
  11. L8:126 excites
  12. L8:958 suppresses
  13. L14:164 suppresses
  14. L6:2112 excites
  15. L11:116 excites
  16. L10:2123 suppresses
  17. L15:195 excites
  18. L3:2034 excites
  19. L10:2633 suppresses
  20. L15:2582 suppresses
  21. L8:2310 suppresses
  22. L12:225 suppresses
  23. L2:1539 suppresses
  24. L10:281 suppresses

Feeds later layers

  1. L25:2055 suppresses
  2. L23:625 suppresses
  3. L19:1424 suppresses
  4. L27:650 suppresses
  5. L20:2059 suppresses
  6. L21:2936 suppresses
  7. L26:2947 suppresses
  8. L24:600 suppresses
  9. L26:846 suppresses
  10. L22:505 suppresses
  11. L21:1892 suppresses
  12. L19:2682 suppresses
  13. L19:2413 weakly suppresses
  14. L23:1807 weakly suppresses
  15. L23:893 weakly suppresses
  16. L27:769 weakly suppresses
  17. L22:477 weakly suppresses
  18. L19:1892 weakly suppresses
  19. L26:1417 weakly suppresses
  20. L26:93 weakly excites
  21. L24:401 weakly excites 'it' referring back to a noun
  22. L24:1080 weakly suppresses
  23. L27:1052 weakly excites
  24. L20:1774 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 -6.86 and -6.512 values between -6.51 and -6.151 values between -6.15 and -5.792 values between -5.79 and -5.431 values between -5.07 and -4.722 values between -4.72 and -4.362 values between -4.36 and -4.002 values between -4.00 and -3.643 values between -3.64 and -3.281 values between -3.28 and -2.933 values between -2.93 and -2.571 values between -2.57 and -2.214 values between -2.21 and -1.857 values between -1.85 and -1.5012 values between -1.50 and -1.1416 values between -1.14 and -0.7845 values between -0.78 and -0.42487 values between -0.42 and -0.062191 values between -0.06 and 0.30932 values between 0.30 and 0.65160 values between 0.65 and 1.0118 values between 1.01 and 1.375 values between 1.37 and 1.731 values between 1.73 and 2.082 values between 2.08 and 2.442 values between 2.80 and 3.162 values between 3.16 and 3.521 values between 4.95 and 5.312 values between 7.10 and 7.45offnegativepositiveoften

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. 某些
  2. 而成
  3. ণ
  4. _than
  5. certain
  6. groups
  7. landers
  8. 他们会
  9. licas
  10. ansion

Suppresses

  1. 都要
  2. 都可以
  3. 都在
  4. 都应该
  5. 都有
  6. 都
  7. 都需要
  8. 都被
  9. 都认为
  10. 均可