Neuron Explorer

Qwen3-0.6B

blank slot before a year or number

A neuron in layer 26 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 empty spot where a number, usually a year, is missing, such as 'since the [blank]1950s'. Turning it up made the model produce strange filler text and repeated words.

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

    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

  2. #2 peak

    investors. This warning contradicts the widely held belief that global crude reserves will remain plentiful for decades.⏎Critics say similar predictions of scarcity at the time of the 1973-74 Arab oil embargo didn’t come true. “

  3. #3 peak

    pain or discomfort, or a severe cough, contact your health care provider, walk-in clinic, or emergency department. If you are having a medical emergency, call 9-1-1.⏎Tips to reduce your smoke exposure⏎Smoke levels may

  4. #4 peak

    . The Washington Post’s Robert Samuelson has done just this in a recent column. As pointed out by Mr. Samuelson blacks have made much progress since the 1960

  5. #5 peak

    /an untrained eye<|endoftext|>From SABR member John Thorn at Our Game on August 18, 2014:⏎When Jackie Robinson opened the 1947

  6. #6 peak

    began a collaboration with DeepMind and the team behind AlphaFold - a machine learning program that attracted significant attention from the structural biology community based on its performance in the 14thCritical Assessment of<|endoftext|>Researchers Show Impaired Muscle Function from Antibacterial

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: (心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先心中先

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

    now accepted defeat.⏎The brief, fragmentary Book <|endoftext|>Welfare state in capitalism⏎Since the eruption of the sub-prime mortgage crises of 2007 to 2009, there have been hot debates on how the

  2. #2 peak

    , fresh produce may be secured. The home vegetable garden is worthy of<|endoftext|>MicroRNAs (miRNAs) are single-stranded, 18–24 nucleotide long, non-coding RNA molecules. They are involved in virtually

  3. #3 peak

    18mm lens (because I wanted to have as much of the sky as possible in my image) my equation was 600/18=33 seconds.⏎With an exposure of 33 seconds I had to have a

  4. #4 peak

    that ignites spontaneously in air when finely divided. Strontium is used in pyrotechnic compounds and various alloys. Atomic number 38; atomic weight 87.62; melting point 76

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. L25:146 strongly excites
  2. L25:854 strongly suppresses
  3. L22:743 strongly suppresses
  4. L21:395 strongly excites
  5. L24:197 strongly excites
  6. L25:232 strongly suppresses
  7. L22:2275 strongly excites
  8. L25:182 strongly excites
  9. L23:199 suppresses
  10. L21:580 suppresses
  11. L25:3007 suppresses the zero digit in decimals
  12. L19:298 excites
  13. L23:416 suppresses
  14. L20:787 excites
  15. L22:193 suppresses
  16. L19:1985 excites
  17. L24:13 excites
  18. L21:110 excites
  19. L22:332 suppresses
  20. L25:95 suppresses
  21. L20:1337 suppresses
  22. L21:339 excites spot just before a year
  23. L24:162 suppresses
  24. L16:372 excites

Feeds later layers

  1. L27:238 strongly suppresses
  2. L27:48 strongly excites
  3. L27:241 strongly excites
  4. L27:283 strongly suppresses the spot right before a number
  5. L27:428 suppresses
  6. L27:530 suppresses
  7. L27:628 suppresses
  8. L27:396 weakly suppresses
  9. L27:328 weakly suppresses
  10. L27:287 weakly suppresses
  11. L27:448 weakly excites
  12. L27:1456 weakly suppresses
  13. L27:132 weakly excites
  14. L27:22 weakly suppresses
  15. L27:1019 weakly suppresses
  16. L27:494 weakly suppresses
  17. L27:1332 weakly suppresses
  18. L27:1132 weakly excites
  19. L27:2249 weakly suppresses
  20. L27:154 weakly excites
  21. L27:913 weakly suppresses
  22. L27:1196 weakly suppresses
  23. L27:323 weakly suppresses
  24. L27:1113 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 -35.91 and -34.411 values between -29.94 and -28.442 values between -25.46 and -23.973 values between -23.97 and -22.474 values between -22.47 and -20.982 values between -20.98 and -19.493 values between -19.49 and -18.004 values between -18.00 and -16.507 values between -16.50 and -15.013 values between -15.01 and -13.522 values between -13.52 and -12.039 values between -12.03 and -10.5315 values between -10.53 and -9.0425 values between -9.04 and -7.55103 values between -7.55 and -6.06211 values between -6.06 and -4.56496 values between -4.56 and -3.07971 values between -3.07 and -1.581187 values between -1.58 and -0.08787 values between -0.08 and 1.4118 values between 1.41 and 2.9015 values between 2.90 and 4.399 values between 4.39 and 5.894 values between 5.89 and 7.387 values between 7.38 and 8.875 values between 8.87 and 10.363 values between 10.36 and 11.863 values between 11.86 and 13.351 values between 13.35 and 14.841 values between 14.84 and 16.332 values between 16.33 and 17.831 values between 17.83 and 19.321 values between 19.32 and 20.812 values between 22.30 and 23.80offnegativepositiveoften

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. ㄒ
  2. Authorities
  3. URLException
  4. "</
  5. URLOPT
  6. StateException
  7. Ḋ
  8. /{{$
  9. HTTPHeader
  10. (${

Suppresses

  1. 1
  2. 2
  3. 3
  4. 5
  5. 8
  6. 4
  7. 9
  8. 6
  9. 0
  10. 7