Neuron Explorer

Qwen3-0.6B

'acid' and related word parts

A neuron in layer 20 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 the word 'acid', especially in words like 'acidophilum', 'acidifies', 'acid rain' and 'acid based'. Turned up, the model produced number lists and repeated years, so the steering output does not show the same 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

    11 November by the Molecular Systems group at the LMS shared insights into these questions.⏎Building chromatin-like structures⏎The archaeon Thermoplasma acidophilum grows optimally in very acidic conditions with high temperatures. It doesn’t

  2. #2 peak

    the usage of plastic bags, plastic straws and switch to a compostable products and reuse those items that are reusable.<|endoftext|>Sixty-three years ago the first acid trip was taken by an unwitting research chemist, Albert Hofmann, who has

  3. #3 peak

    hydrologic consideration.⏎—Pervious concrete, water purification, acid rain, waste motor oil.⏎Ming-Ju Lee is with the Jen-T

  4. #4 peak

    we’ve harvested.⏎Now it’s time to go plant a reef.⏎I got here with a question: What happens to coral reefs as the ocean warms and acidifies over the coming century? Corals might seem removed from human society, but

  5. #5 peak

    's has introduced several preprocessing methods to reduce the log data size and to eliminate incorrect records, which lead to improved quality, quantity and performance.<|endoftext|>Based on the acid rain and concurrent meteorological observational data during the past 10 years in Fuj

  6. #6 peak

    Qualitative Crackle test is used for engine oils.)⏎- Acid Number measures acid buildup, which denotes oxidation.⏎- FTIR measures chemistry changes in a lubricant

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 first check to check the first check, the first check to check the first check... The first check to check the first check, the first check to check the first check... I am aware that the first check check check check check check

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

    secure digital communications. Specifically, if large quantum computers can be built then they may be able to break the asymmetric cryptographic primitives that are currently used in TLS, the security protocol behind HTTPS.⏎Quantum computers exist today but, for the moment, they

  2. #2 peak

    , without safety screening⏎Reproduced with permission from Fluechtling.net⏎- Millions

  3. #3 peak

    the adhesive or sealant to take advantage of the microscopic roughness etched to the surface by atmospheric plasma, locking them together. In<|endoftext|>Here at CRC, safety is our number one priority. We have dedicated staff who can offer a range of

  4. #4 peak

    advances in a decade.⏎Rechargeable lithium-ion batteries became costly for General Motors Co. (GM) after they caught fire in its plug-in Volt car during safety tests in 2011. They grounded Boeing Co. (BA)

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. L19:407 excites
  2. L4:1761 suppresses
  3. L19:314 excites
  4. L18:287 excites
  5. L4:2347 weakly suppresses
  6. L2:852 weakly suppresses
  7. L4:741 weakly excites
  8. L4:834 weakly suppresses
  9. L5:800 weakly suppresses
  10. L0:2907 weakly suppresses
  11. L4:428 weakly suppresses
  12. L0:2468 weakly suppresses
  13. L19:256 weakly suppresses
  14. L17:650 weakly excites
  15. L18:519 weakly excites
  16. L5:1946 weakly suppresses
  17. L18:1069 weakly excites
  18. L8:500 weakly excites 'to' after verbs like continue or help
  19. L6:105 weakly excites
  20. L13:743 weakly suppresses
  21. L10:299 weakly suppresses
  22. L0:3031 weakly suppresses
  23. L18:2360 weakly suppresses
  24. L15:2804 weakly excites

Feeds later layers

  1. L23:670 suppresses
  2. L25:1401 weakly excites
  3. L25:1890 weakly suppresses
  4. L24:2737 weakly excites
  5. L22:2584 weakly suppresses
  6. L24:886 weakly excites
  7. L27:2375 weakly suppresses
  8. L24:1112 weakly suppresses
  9. L22:775 weakly suppresses
  10. L24:1775 weakly suppresses
  11. L21:1116 weakly suppresses
  12. L23:808 weakly suppresses
  13. L24:2500 weakly suppresses
  14. L21:1887 weakly suppresses
  15. L22:229 weakly suppresses "of" after a noun in "X of Y"
  16. L22:2477 weakly excites
  17. L21:2535 weakly excites
  18. L21:2814 weakly suppresses
  19. L24:841 weakly suppresses
  20. L23:996 weakly suppresses
  21. L23:1227 weakly suppresses
  22. L23:2081 weakly suppresses
  23. L27:2725 weakly excites
  24. L23:1592 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 -47.03 and -45.641 values between -40.05 and -38.651 values between -38.65 and -37.261 values between -33.07 and -31.673 values between -23.29 and -21.901 values between -21.90 and -20.502 values between -19.10 and -17.711 values between -17.71 and -16.311 values between -16.31 and -14.911 values between -14.91 and -13.525 values between -13.52 and -12.123 values between -12.12 and -10.7210 values between -10.72 and -9.337 values between -9.33 and -7.934 values between -7.93 and -6.538 values between -6.53 and -5.1423 values between -5.14 and -3.7432 values between -3.74 and -2.34110 values between -2.34 and -0.953277 values between -0.95 and 0.45312 values between 0.45 and 1.8571 values between 1.85 and 3.2420 values between 3.24 and 4.647 values between 4.64 and 6.042 values between 6.04 and 7.434 values between 7.43 and 8.83offnegativepositiveoften

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. inea
  2. аль
  3. igm
  4. ighet
  5. ifiers
  6. gency
  7. alendar
  8. ifferential
  9. als
  10. äl

Suppresses

  1. 系数
  2. �数
  3. preocup
  4. HECK
  5. scares
  6. 饵
  7. checks
  8. ECH
  9. vos
  10. poids