Neuron Explorer

Qwen3-0.6B

mid-word fragments in capitalised titles

A neuron in layer 20 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the piece of a word that comes in the middle of a longer word, especially in all-caps headings and titles. It is hard to say what it does from the examples, and the steering outputs did not show a clear theme.

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

    eth not at all. '⏎AKAI HEIRLOOM RICE - GRAIN AKAI RICE is a heirloom rice

  2. #2 peak

    of Alzheimer’s.⏎Courtesy of Baycrest, the global leader in innovations, research and breakthroughs for the journey of aging.⏎ALSO ON EVERYTHINGZOOMER:<|endoftext|>SeaWiFS: Gulf Coast Sediments⏎Provided by the Sea

  3. #3 peak

    , and English character and manners described.?1 Paradoxically, these three editions?The Juvenile<|endoftext|>CASE NUMBER: 212⏎CASE MNEMONIC: BASMEX⏎CASE NAME: Mexico and Basel⏎On the Mexico-US

  4. #4 peak

    you like to know what each chakra represents? Here is a simple diagram and list for you to follow. | WIKIPEDIA⏎THE CHAKRAS ARE AS FOLLOWS⏎- ROOT/BASE (MULADH

  5. #5 peak

    6⏎THE CHAKRAS AND CHAKRA BALANCING MEDITATION⏎SPINNING WHEELS (SPHERES) OF LIGHT CODES AND PROGRAMS⏎Vetruvian Man with Seven Chakra Overlay⏎

  6. #6 peak

    largest wetland in the North Island and is recognised as a Ramsar site – a wetland of international importance.⏎Ō Tū WharekaiCANTERBURY Ō Tū Wharekai is one of the best intact

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: **A **C** -F (C, F) (C, F)** **C** -**F** (C, F) **C** -**F** (C, F) **C** (

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

    |⏎|Number of pages||12|⏎|Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most

  2. #2 peak

    single-layer coated, triple-layer coated and graded-index-layer coated samples was performed and compared with experimentally derived films. Good agreement between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Cer

  3. #3 peak

    work in the jungle, there’s nothing Milton likes better than fresh fruit bat soup.⏎IS39 is a seven-element array deployed in a triangular geometry with four elements in

  4. #4 peak

    .⏎Are standards mandatory ?⏎Standards are considered voluntary because they serve as guidelines, but do not of themselves have the force of law. ASME cannot force any manufacturer, inspector, or installer to follow ASME standards.

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. L18:373 strongly suppresses
  2. L16:584 strongly excites
  3. L19:1731 strongly suppresses
  4. L5:35 strongly suppresses
  5. L18:599 suppresses
  6. L10:115 excites the word COVID before -19
  7. L4:533 suppresses
  8. L6:2269 suppresses
  9. L17:699 excites
  10. L1:2221 excites
  11. L16:158 suppresses
  12. L3:627 suppresses
  13. L0:76 excites
  14. L19:188 excites
  15. L15:257 suppresses
  16. L18:546 excites
  17. L7:128 suppresses
  18. L17:1747 suppresses split tech names and acronyms
  19. L19:2344 suppresses
  20. L6:430 suppresses
  21. L10:2313 excites
  22. L19:712 suppresses
  23. L3:1479 suppresses
  24. L0:1904 suppresses

Feeds later layers

  1. L22:651 strongly suppresses
  2. L22:1126 strongly suppresses
  3. L26:1531 strongly suppresses
  4. L27:1360 strongly suppresses
  5. L24:980 strongly suppresses
  6. L24:1991 strongly suppresses
  7. L21:533 strongly suppresses
  8. L27:1879 strongly suppresses
  9. L25:1750 strongly suppresses
  10. L26:434 strongly suppresses
  11. L22:212 strongly suppresses
  12. L26:2849 suppresses
  13. L24:1970 suppresses
  14. L27:1520 suppresses
  15. L27:2146 suppresses
  16. L27:782 suppresses
  17. L25:244 excites
  18. L27:2505 suppresses
  19. L26:155 excites
  20. L27:1906 suppresses
  21. L27:491 excites
  22. L24:938 excites
  23. L27:1685 suppresses
  24. L26:816 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 -20.31 and -19.453 values between -17.74 and -16.881 values between -14.30 and -13.443 values between -13.44 and -12.591 values between -12.59 and -11.731 values between -11.73 and -10.872 values between -10.01 and -9.152 values between -9.15 and -8.293 values between -6.58 and -5.722 values between -5.72 and -4.861 values between -4.86 and -4.004 values between -4.00 and -3.143 values between -3.14 and -2.282 values between -2.28 and -1.421 values between -1.42 and -0.571710 values between -0.57 and 0.291988 values between 0.29 and 1.15167 values between 1.15 and 2.018 values between 2.01 and 2.871 values between 3.73 and 4.591 values between 6.30 and 7.161 values between 7.16 and 8.021 values between 8.88 and 9.741 values between 13.17 and 14.03offnegativepositiveoften

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. ours
  2. ourselves
  3. uellement
  4. SqlServer
  5. hers
  6. Eb
  7. oneself
  8. herself
  9. angep
  10. 该项

Suppresses

  1. s
  2. ING
  3. AND
  4. ICS
  5. ANT
  6. 必
  7. ABLE
  8. ORT
  9. ER
  10. OR