Neuron Explorer

Qwen3-0.6B

opening parenthesis starting an aside

A neuron in layer 13 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 fires on an opening bracket that starts an explanation or example right after a noun or phrase. When it was turned up, the model's text got stuck repeating words instead of making sense.

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

    volume of the container it is in. Despite all of the constituent particles being charged, typically the plasma itself has no overall charge. However, some plasmas (non-neutral) can be created with an overall charge (either positive or negative)

  2. #2 peak

    carcinomatosis—are cancers that are frequently referred to as peritoneal cancers. Whether peritoneal cancer starts in the peritoneum or spreads from somewhere else (colon⏎Ovarian Epithelial, Fallopian Tube, and Primary

  3. #3 peak

    enjoy their colorful visuals, whimsical themes, and, in some cases, varied functions (for instance, pop-up, or moving-part valentines). Displayed singly

  4. #4 peak

    ). This may be further exacerbated by few harmful chemicals (proteins) such as tumor necrosis factor alpha (TNF-α

  5. #5 peak

    a savings account.⏎Once the “long-term savings” envelope reaches a certain dollar amount (you can agree upon this amount with your child beforehand), it’s time to open

  6. #6 peak

    the plasma conversion step started with two pilot plants: the first, to test the gasification and plasma conversion; the second, to test the methanation process (where

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 day I was going to go to the sun (I'm going to go to the sun (I'm going to go to the sun (I'm going to go to the sun (I'm going to go to the sun (I

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

    is Noether's theorem, which states that there is a one-to-one correspondence between each one of them and a differentiable symmetry of nature. For example,

  2. #2 peak

    of the operating principle of relationship between components. Also knows as a Schematic or Diagrammatic Drawing.⏎13. Perspective Drawing⏎Descriptive three-quarter view produced using a perspective drawing technique. Created using line only without the application of color or

  3. #3 peak

    System (DNS) is a global database that translates domain names (such as www.internetsociety.org) to Internet addresses that

  4. #4 peak

    amayani (Hindi:कामायनी) (1936) is a Hindi epic poem (Mahakavya) by Jaishankar Prasad (1889–1937

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. L9:1414 suppresses opening parenthesis after copyright or year
  2. L5:330 excites opening parenthesis after a noun
  3. L7:110 weakly excites
  4. L9:1370 weakly suppresses
  5. L8:2184 weakly excites
  6. L11:2400 weakly excites
  7. L12:2654 weakly suppresses
  8. L10:270 weakly excites
  9. L6:1391 weakly suppresses
  10. L11:257 weakly excites
  11. L6:304 weakly excites
  12. L10:1756 weakly excites
  13. L12:1557 weakly suppresses comma after an introductory phrase
  14. L12:1248 weakly excites
  15. L12:878 weakly suppresses
  16. L7:25 weakly excites
  17. L8:81 weakly suppresses
  18. L5:2548 weakly excites
  19. L2:1982 weakly suppresses
  20. L4:2540 weakly excites
  21. L11:1855 weakly suppresses
  22. L2:2703 weakly suppresses
  23. L7:2239 weakly excites
  24. L10:166 weakly suppresses

Feeds later layers

  1. L14:1616 suppresses
  2. L16:1643 suppresses
  3. L16:530 suppresses
  4. L15:113 excites
  5. L15:2956 weakly excites
  6. L15:2222 weakly suppresses
  7. L16:1128 weakly suppresses
  8. L16:2093 weakly suppresses 'of' in 'piece of' phrases
  9. L17:1229 weakly excites
  10. L16:2705 weakly suppresses
  11. L15:1528 weakly excites
  12. L15:261 weakly excites
  13. L20:1012 weakly suppresses
  14. L18:3033 weakly excites
  15. L16:659 weakly excites
  16. L14:2797 weakly suppresses
  17. L15:497 weakly suppresses
  18. L15:1156 weakly excites
  19. L14:1928 weakly suppresses
  20. L16:1394 weakly suppresses
  21. L25:880 weakly excites comma after a cause or effect clause
  22. L17:1516 weakly suppresses
  23. L15:2928 weakly suppresses
  24. L14:2022 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.

6 values between -1.75 and -1.553 values between -1.55 and -1.353 values between -1.35 and -1.1515 values between -1.15 and -0.9559 values between -0.95 and -0.7597 values between -0.75 and -0.55231 values between -0.55 and -0.35522 values between -0.35 and -0.151621 values between -0.15 and 0.06829 values between 0.06 and 0.26285 values between 0.26 and 0.4694 values between 0.46 and 0.6646 values between 0.66 and 0.8623 values between 0.86 and 1.0611 values between 1.06 and 1.2612 values between 1.26 and 1.469 values between 1.46 and 1.6610 values between 1.66 and 1.869 values between 1.86 and 2.067 values between 2.06 and 2.263 values between 2.26 and 2.461 values between 2.66 and 2.862 values between 3.26 and 3.461 values between 4.46 and 4.661 values between 4.86 and 5.062 values between 5.06 and 5.263 values between 5.26 and 5.461 values between 5.87 and 6.072 values between 6.07 and 6.27offnegativepositiveoften

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. inee
  2. den
  3. potentially
  4. possibly
  5. aghan
  6. chool
  7. 偓
  8. xm
  9. brace
  10. _MAY

Suppresses

  1. dern
  2. perimental
  3. /Application
  4. 迸
  5. olatile
  6. نتج
  7. .getFont
  8. .FONT
  9. 纯
  10. ,true