Neuron Explorer

Qwen3-0.6B

the gap right before a number

A neuron in layer 24 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

The neuron fires at a spot in the text where a number seems to be missing or was stripped out, such as between 'Hebrews' and '9:22' or 'equal to' and '252 calories'. Steering did not give a clear picture: the outputs drifted into unrelated text, commas, or brackets.

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

    law purged with blood, and without shedding of blood is no remission” (Hebrews 9:22). To purge is to “get rid of whatever is imp

  2. #2 peak

    L. Three Special Limits 1. sin lim 1 x x x = 2. 1 cos lim x x x -= 3. ( ) 1 lim 1 x x x e + = Strategy for Finding Limits:

  3. #3 peak

    .⏎- If we want to write this in differential form, which is useful when we do our u-substitution,⏎- we could say that du is equal to 5 cosine 5x.⏎- Now when you look over here, we don

  4. #4 peak

    hours of high-definition video in about a cup of DNA.⏎There is a lot of digital information in the world about three zettabytes' worth (that's 3000 billion billion bytes) and the constant influx of new digital content

  5. #5 peak

    battery is intended.⏎Btu (British Thermal Unit) — The amount of heat required to raise the temperature of one pound of water one degree Fahrenheit; equal to 252 calories.⏎Charge Controller — A component of a phot

  6. #6 peak

    du could be equal to -- so du/dx (derivative of u with respect to x),⏎- well we just use the chain rule. Derivative of 5x is 5

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: when I read a book that has a very long text, they often use several symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called symbols called

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

    cultural and economic forces combine to create a 24-hour society in which millions of Americans—either chronically or intermittently—obtain insufficient sleep as a result of workplace and lifestyle determinants. A convincing

  2. #2 peak

    of name and historical aspects⏎Erica plukenetii was named by Linnaeus in honour of Leonard Plukenet (1641-1706), who was a physician by profession but also a botan

  3. #3 peak

    espasianus Augustus, original name Titus Flavius Vespasianus, (born Dec. 30, 39 ce—died Sept.

  4. #4 peak

    , Sisley), he attended from 1875 training in the painter-glassblower Alexandre Mauvernay workshop (1810-1898) who worked in

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. L9:7 excites
  2. L21:173 excites
  3. L20:563 excites
  4. L18:48 suppresses
  5. L14:348 suppresses
  6. L20:499 excites
  7. L17:38 suppresses
  8. L18:93 suppresses
  9. L16:527 suppresses
  10. L22:424 suppresses
  11. L23:80 excites blank slot before a year or number
  12. L15:1601 weakly excites
  13. L10:33 weakly suppresses
  14. L14:1091 weakly excites
  15. L19:131 weakly suppresses
  16. L21:390 weakly suppresses
  17. L1:8 weakly excites
  18. L14:51 weakly excites colon that introduces a statement
  19. L23:262 weakly suppresses
  20. L19:119 weakly suppresses
  21. L15:44 weakly suppresses
  22. L15:123 weakly suppresses
  23. L14:775 weakly suppresses
  24. L5:5 weakly suppresses digit after a label like paper or type

Feeds later layers

  1. L25:239 suppresses
  2. L26:17 suppresses
  3. L25:15 excites
  4. L26:56 suppresses
  5. L25:146 suppresses
  6. L25:222 suppresses
  7. L27:627 suppresses
  8. L26:21 suppresses
  9. L27:448 suppresses
  10. L26:199 suppresses
  11. L26:254 suppresses
  12. L27:558 suppresses
  13. L26:263 suppresses
  14. L26:48 excites
  15. L26:503 weakly suppresses
  16. L25:95 weakly excites
  17. L26:352 weakly suppresses
  18. L27:145 weakly excites dots inside web addresses
  19. L27:83 weakly suppresses
  20. L25:30 weakly excites
  21. L27:732 weakly suppresses
  22. L25:232 weakly excites
  23. L27:817 weakly suppresses
  24. L27:184 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 -29.50 and -28.372 values between -23.86 and -22.733 values between -22.73 and -21.601 values between -21.60 and -20.484 values between -20.48 and -19.358 values between -19.35 and -18.222 values between -18.22 and -17.096 values between -17.09 and -15.969 values between -15.96 and -14.832 values between -14.83 and -13.718 values between -13.71 and -12.582 values between -12.58 and -11.4513 values between -11.45 and -10.3226 values between -10.32 and -9.1953 values between -9.19 and -8.0778 values between -8.07 and -6.94147 values between -6.94 and -5.81226 values between -5.81 and -4.68259 values between -4.68 and -3.55369 values between -3.55 and -2.43518 values between -2.43 and -1.30825 values between -1.30 and -0.171068 values between -0.17 and 0.96154 values between 0.96 and 2.0960 values between 2.09 and 3.2221 values between 3.22 and 4.3422 values between 4.34 and 5.4710 values between 5.47 and 6.605 values between 6.60 and 7.734 values between 7.73 and 8.861 values between 12.24 and 13.371 values between 14.50 and 15.63offnegativepositiveoften

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. -Mart
  2. -script
  3. -ing
  4. -note
  5. £o
  6. -fashion
  7. -ce
  8. -db
  9. -ground
  10. ­i

Suppresses

  1. ris
  2. respons
  3. vis
  4. kills
  5. skills
  6. weighted
  7. отнош
  8. strings
  9. CString
  10. guided