Neuron Explorer

Qwen3-0.6B

reflect and reflection words

A neuron in layer 23 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 reacts to words about light bouncing off surfaces, like 'reflect', 'reflected', 'reflective' and 'mirror'. When it was turned up, the model wrote about reflections and kept repeating the word 'reflection'.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    using a Cannon SLR camera and the expertise of our instructor, Sallyanne.⏎As we have explored the way light works we’ve seen how light can refract, reflect and be manipulated through translucent or opaque objects. With this in mind the boys

  2. #2 peak

    μm. The absorptivity and emissivity around freezing point is 0.99 – meaning that only 1% of incident longwave radiation would be reflected.⏎Let’s take a look at the Planck curve – the blackbody

  3. #3 peak

    arrange for special meetings outside of school hours. You may<|endoftext|>Definitions of focus⏎n. - A point in which the rays of light meet, after being reflected or refrcted, and at which the image is formed; as, the focus

  4. #4 peak

    which photon emission is organized. Main parts of laser are power supply, lasing medium and a pair of precisely aligned mirrors. One has totally reflective surface and other is partially reflective (96 %). The most important part of laser apparatus is laser

  5. #5 peak

    missivity is the ability of a material to radiate energy. When heat or light energy—typically from the sun or HVAC system—is absorbed by glass it is either shifted away by air movement or re-radiated by the glass surface.⏎Typ

  6. #6 peak

    ⏎solid, liquid and gaseous forms respectively.⏎Physical Properties of Minerals⏎Minerals can be identified by their hardness,⏎colour, the way they reflect light (lustre), the⏎way they break (cleavage and fracture), and⏎

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 two main things that would be key to a successful, competent, and competent team as well as a competent and competent team would be the two key things, and the two key things are the same thing. Okay, okay, so

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

    the desert air.⏎So it appears that they have been homophones, at least in the London dialect, for over four hundred years.⏎Wordsworth (1770–1850) also appears to have only used the one

  2. #2 peak

    - IUCN Status:Least Concern<|endoftext|>Introduction to Lead Screws⏎Lead Screws are remarkably common components, and can be found in engineering environments operating lathes, CNC machines or complex robotics, but they can just as commonly be found

  3. #3 peak

    ray passes from a substance of one density into a substance of different density, the speed of light changes and bending of light occurs.⏎In air, light travels at 3 x 108 m/sec but in water, it travels only about

  4. #4 peak

    æ are often called aphis-lions. Also called golden-eyed fly.⏎- n. Melithreptus lunulatus, a small species of honey-eater.⏎GNU Webster's 1

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. L20:2287 suppresses
  2. L22:2259 excites
  3. L21:3067 weakly suppresses
  4. L22:2422 weakly excites
  5. L4:2849 weakly excites
  6. L10:1551 weakly excites
  7. L21:2965 weakly excites
  8. L10:780 weakly excites
  9. L4:2065 weakly excites
  10. L21:1630 weakly excites
  11. L4:1531 weakly excites
  12. L20:2709 weakly suppresses
  13. L0:3071 weakly suppresses
  14. L4:2390 weakly suppresses
  15. L9:2719 weakly excites
  16. L21:1015 weakly excites
  17. L22:2022 weakly suppresses
  18. L22:1746 weakly excites
  19. L6:2107 weakly suppresses
  20. L10:670 weakly suppresses
  21. L8:978 weakly suppresses
  22. L11:1940 weakly suppresses
  23. L3:2609 weakly suppresses
  24. L10:440 weakly excites

Feeds later layers

  1. L27:1531 strongly excites
  2. L25:2593 excites
  3. L26:1559 excites
  4. L25:1803 excites
  5. L24:157 weakly suppresses
  6. L26:2586 weakly excites
  7. L25:3066 weakly excites
  8. L25:2384 weakly excites
  9. L25:1494 weakly excites
  10. L25:2495 weakly excites
  11. L24:160 weakly excites
  12. L25:923 weakly suppresses
  13. L25:2395 weakly excites
  14. L24:2169 weakly excites
  15. L25:534 weakly excites
  16. L26:2349 weakly excites
  17. L24:852 weakly excites
  18. L24:2883 weakly excites
  19. L24:1360 weakly excites
  20. L24:1301 weakly excites
  21. L25:2525 weakly excites
  22. L24:836 weakly suppresses
  23. L24:259 weakly excites
  24. L25:1856 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 -12.44 and -10.451 values between -10.45 and -8.462 values between -8.46 and -6.488 values between -6.48 and -4.499 values between -4.49 and -2.50226 values between -2.50 and -0.513566 values between -0.51 and 1.4843 values between 1.48 and 3.4623 values between 3.46 and 5.458 values between 5.45 and 7.448 values between 7.44 and 9.432 values between 9.43 and 11.414 values between 11.41 and 13.402 values between 13.40 and 15.391 values between 19.36 and 21.351 values between 31.29 and 33.281 values between 45.20 and 47.191 values between 47.19 and 49.181 values between 65.08 and 67.06offnegativepositiveoften

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. reflect
  2. reflecting
  3. reflection
  4. reflective
  5. reflects
  6. 反射
  7. reflected
  8. Reflect
  9. Reflect
  10. reflections

Suppresses

  1. competency
  2. compet
  3. itzer
  4. 游
  5. 敦
  6. dragon
  7. 벨
  8. ERY
  9. iddy
  10. uffs