Neuron Explorer

Qwen3-0.6B

parentheses and code fragments

A neuron in layer 11 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 responds to opening brackets and punctuation in technical text, such as '(CDC)' after an organization name and parts of code. It is mixed, and the strongest hit is a bullet mark in a forum post. Turning it up or down did not give a clear story.

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

    in the field from August 2 – August 20.<|endoftext|>Genetics as it applies to evolution, molecular biology, and medical aspects.⏎10 posts • Page 1 of 1⏎Hope somebody can help, I've got an

  2. #2 peak

    11). http://www.cdc<|endoftext|>|This article is about the assassin of pharaoh Cleopatra. For other uses, see Aya or Amun.|⏎Prior to establishing the Hidden Ones, A

  3. #3 peak

    age of 25 in 2006.⏎Researchers T.J. Mathews and Brady E. Hamilton of the Center for Disease Control and Prevention’s (CDC) National Center for Health Statistics, Division of Vital Statistics, Reproductive

  4. #4 peak

    a charge of bad taste,<|endoftext|>8 edition of Principles of Biblical Interpretation found in the catalog.⏎April 1, 2

  5. #5 peak

    Released: Nov 9, 1993.⏎- Full Report:⏎Pursuant to a congressional request, GAO reviewed the Department of Energy's (DOE) plans

  6. #6 peak

    , December 2, 2012.⏎The contest is a partnership with The Planetary Society in Pasadena, California; the Massachusetts Institute of Technology’s (MIT) Lincoln Laboratory in Lexington, Massachusetts; and the University of Arizona in Tucson

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 am not going to be in a position of need to use my resources for the benefit of a possible crisis of need..." I am not going to be in a position of need to use my resources for the benefit of a possible crisis

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

    -- Important numbers!⏎xs = [6, 9, 13, 42]⏎-- Without library⏎f :: Bool -> [Int]⏎f b = if b then xs else ⏎-- With library⏎g :: Bool -> [

  2. #2 peak

    found by quadrature and is⏎To use MATLAB we first construct an M-file that defines the equation.⏎% filename rhs.m⏎function ydot=rhs(t,y)%compute the rhs of eqns.⏎%for any input values t, y

  3. #3 peak

    parent graph of the cotangent function f(x) = cot x,⏎by using the same techniques you use to find the tangent

  4. #4 peak

    conuts and banana leaves<|endoftext|>const void * memchr ( const void * ptr, int value, size_t num );⏎void * memchr ( void * ptr, int value, size_t num );⏎Locate character in block of memory⏎

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. L8:1610 suppresses
  2. L10:406 excites
  3. L10:392 excites
  4. L5:272 excites
  5. L10:1187 excites
  6. L7:248 excites
  7. L6:276 suppresses
  8. L8:288 suppresses
  9. L7:457 excites
  10. L1:993 weakly suppresses
  11. L2:296 weakly suppresses
  12. L10:779 weakly suppresses
  13. L10:1425 weakly excites
  14. L7:718 weakly excites
  15. L7:513 weakly excites
  16. L8:1334 weakly suppresses
  17. L10:1251 weakly suppresses
  18. L10:1625 weakly suppresses
  19. L3:1634 weakly excites
  20. L7:158 weakly excites
  21. L8:144 weakly suppresses
  22. L4:2907 weakly excites
  23. L1:1535 weakly excites
  24. L10:681 weakly suppresses

Feeds later layers

  1. L15:1775 suppresses
  2. L16:1856 suppresses
  3. L12:1163 suppresses
  4. L16:2881 suppresses
  5. L12:855 suppresses
  6. L15:2582 suppresses
  7. L15:2454 suppresses
  8. L15:487 suppresses
  9. L13:568 suppresses
  10. L18:653 suppresses
  11. L12:829 suppresses
  12. L13:1463 suppresses
  13. L13:1987 suppresses
  14. L17:946 suppresses
  15. L13:711 suppresses
  16. L16:375 suppresses
  17. L14:726 suppresses
  18. L21:1347 weakly suppresses
  19. L13:2612 weakly suppresses
  20. L13:845 weakly suppresses
  21. L12:1298 weakly suppresses
  22. L16:1402 weakly suppresses
  23. L13:1283 weakly suppresses
  24. L12:443 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 -1.20 and -1.131 values between -0.54 and -0.482 values between -0.48 and -0.414 values between -0.41 and -0.3510 values between -0.35 and -0.2822 values between -0.28 and -0.2244 values between -0.22 and -0.15128 values between -0.15 and -0.08292 values between -0.08 and -0.02705 values between -0.02 and 0.05987 values between 0.05 and 0.11794 values between 0.11 and 0.18487 values between 0.18 and 0.24255 values between 0.24 and 0.31119 values between 0.31 and 0.3733 values between 0.37 and 0.4418 values between 0.44 and 0.514 values between 0.51 and 0.571 values between 0.77 and 0.831 values between 1.36 and 1.42offnegativepositiveoften

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. .side
  2. side
  3. idence
  4. stown
  5. continent
  6. 高的
  7. aan
  8. RoleId
  9. estone
  10. missible

Suppresses

  1. 濯
  2. ilos
  3. REST
  4. rest
  5. proj
  6. ldr
  7. /array
  8. 项目的
  9. 援
  10. 缇