Neuron Explorer

Qwen3-0.6B

'from' after movement or transfer verbs

A neuron in layer 10 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 'from' when it describes moving or passing from one place or thing to another, like 'jumps from one surface' or 'transfer of control from the robot'. It goes strongly negative on 'from' in other senses, such as 'art from this artist' or 'get enough from food'. Turning it up did not give a clear 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

    dyslexia for both linguistic and nonlinguistic domains, but speech-specific adaptation deficits. Finally, mediation models in children and adults revealed that the causal pathways from basic perception and adaptation to phonological awareness through speech categorization were not significant.

  2. #2 peak

    off a ledge, for instance -- unless a human needs the robot to drive off the ledge. When those situations happen, you need to have smooth transfer of control from the robot to the appropriate human,” Woods said.⏎“The bottom line is, robots

  3. #3 peak

    han, who conducted the research while at Villanova University in Villanova, Pennsylvania.⏎Yet the fast, stealthy Bagheera has figured out how to leap from thorn to thorn to collect its meal—while avoiding the highly aggressive ants

  4. #4 peak

    the direction) of a wave (light or sound). When a light ray passes from one transparent medium to another, it changes speed and bends.⏎- Light traveling from air and going into water – half submerged straight stick or pencil into water. The

  5. #5 peak

    Do not freeze. Discard if the vaccine has been frozen.⏎A separate sterile syringe and needle must be used for each injection to prevent transmission of infectious agents from one person to another. Needles should be disposed of properly and not recapped

  6. #6 peak

    .⏎The spinal cord runs through the spinal canal, a tunnel that is formed by the holes in the center of each of the vertebrae. Nerves branch from the spinal cord, pass through openings (foraminae) between the vertebra

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: today the stars of a real life (or everyday experiences) (the things around the outside world, the history of science and technology, and the knowledge of the natural world (physics of nature, or mathematics (or mathematics of experience, or

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

    finished project! I can’t wait to have her create more pieces of art from this famous artist in this book and from the other books by the authors.<|endoftext|>

  2. #2 peak

    promise of protecting bones and preventing bone diseases such as osteoporosis. Supplemental vitamin D is popular because it is difficult (if not impossible) to get enough from food. In addition, as noted by the Centers for Disease Control and Prevention (

  3. #3 peak

    del Molino as well.⏎Live the life of a professional archaeologist as you unearth artifacts, then clean and study them. You might also try rebuilding original objects from the fragments you’ve discovered. You’ll have an unparalleled chance to get close to

  4. #4 peak

    2D plans.⏎All the codes and supporting files for this course are available at - https://github.com/PacktPublishing/Create-photorealistic-house-from-scratch-to-end-in-AutoCAD⏎- Publication date:⏎- September

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. L4:192 excites
  2. L5:2925 excites
  3. L9:1382 suppresses
  4. L6:2527 excites
  5. L4:1394 excites
  6. L4:1151 excites
  7. L2:1848 suppresses
  8. L5:1922 suppresses 'to' in 'X to Y' phrases
  9. L6:1726 suppresses
  10. L4:2269 suppresses
  11. L0:1660 weakly suppresses
  12. L6:2937 weakly excites 'by' after a title or author credit
  13. L9:723 weakly suppresses
  14. L9:2451 weakly excites 'for' giving a reason or cause
  15. L9:1340 weakly suppresses
  16. L8:1627 weakly suppresses
  17. L6:2567 weakly suppresses "by" after a passive verb
  18. L4:2347 weakly excites
  19. L4:1122 weakly excites
  20. L4:1126 weakly excites
  21. L6:1997 weakly suppresses
  22. L9:1465 weakly excites
  23. L2:1857 weakly suppresses
  24. L8:2842 weakly excites

Feeds later layers

  1. L17:1721 suppresses 'by' introducing a method or means
  2. L15:2443 suppresses
  3. L18:2916 suppresses
  4. L12:1137 suppresses
  5. L13:831 suppresses
  6. L14:2972 excites
  7. L15:1642 suppresses
  8. L12:2685 excites
  9. L13:2832 suppresses "by" after a passive verb
  10. L19:2273 suppresses
  11. L15:1386 suppresses prepositions after passive verbs
  12. L14:885 suppresses
  13. L13:2509 suppresses
  14. L14:709 suppresses
  15. L12:2888 suppresses
  16. L19:2620 weakly suppresses
  17. L11:2158 weakly suppresses
  18. L14:2180 weakly suppresses
  19. L17:2621 weakly suppresses
  20. L17:2145 weakly suppresses
  21. L17:1058 weakly suppresses
  22. L17:1801 weakly suppresses
  23. L19:1971 weakly suppresses
  24. L18:1235 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 -6.50 and -6.261 values between -3.92 and -3.691 values between -3.69 and -3.461 values between -3.22 and -2.993 values between -2.99 and -2.763 values between -2.76 and -2.523 values between -2.52 and -2.291 values between -2.29 and -2.051 values between -2.05 and -1.823 values between -1.82 and -1.593 values between -1.59 and -1.353 values between -1.35 and -1.129 values between -1.12 and -0.8820 values between -0.88 and -0.6542 values between -0.65 and -0.42149 values between -0.42 and -0.182117 values between -0.18 and 0.051382 values between 0.05 and 0.29140 values between 0.29 and 0.5215 values between 0.52 and 0.758 values between 0.75 and 0.991 values between 1.22 and 1.451 values between 2.62 and 2.86offnegativepositiveoften

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. 僇
  2. poses
  3. .getAmount
  4. .getB
  5. AppState
  6. ooting
  7. posX
  8. aversable
  9. .addData
  10. arded

Suppresses

  1. incer
  2. 前辈
  3. coles
  4. ------⏎⏎
  5. ặt
  6. փ
  7. ми
  8. 缘
  9. entifier
  10. force