Neuron Explorer

Qwen3-0.6B

"layer" in thin layer or ozone layer

A neuron in layer 5 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 "layer," such as in "thin layer" or "ozone layer." The strong examples come from science and everyday descriptions of materials. Turning it up did not produce any clear theme in the model's replies.

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

    work function is very low. An electron can easily cross, out of the material an into the silicon. Electronics is in immediate contact with magnetics.⏎The thin layer of gadolinium is created via an evaporation process, enabling varying of the

  2. #2 peak

    80nm and is therefore invisible to the naked eye. Naturally occurring UV-C rays from the sun are prevented from reaching the earth’s surface by the Ozone layer. However, they can be generated

  3. #3 peak

    .⏎Tools and Resources for App Coding⏎To facilitate the process of coding for<|endoftext|>A sprayed seal is a thin layer of binder sprayed onto a pavement surface with a layer of aggregate incorporated and which is

  4. #4 peak

    to life on Earth.⏎The scientists announced their results at the Australian Meteorological and Oceanographic Society conference at the Australian National University in Canberra last Friday .⏎The ozone layer is in the stratosphere, the zone 10km to 5

  5. #5 peak

    global warming and the thinning of the ozone layer. Ways to overcome this problem are filtration of smoke before being released into the environment

  6. #6 peak

    The Sun's chromosphere is a faint layer of gas surrounding the photosphere, its visible surface. The chromosphere glows crimson at a wavelength of 656.28 nanometers

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: In the case of the question, if you have a problem, you will find the solution. A problem. But no information. So, this is a problem. Wait, this is the problem. I don't have that.

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

    harmless.⏎UV offers a distinct advantage over chlorine-based disinfection because of its ability to inactivate protozoa that threaten public health – such as Cryptosporidium and Giardia – making UV disinfection an

  2. #2 peak

    a phenomenon; a motive for an action or a determination; proof, more or less decisive, for an opinion or a conclusion; principle; efficient cause; final cause; ground of argument.⏎2. (n.) The faculty or capacity of the

  3. #3 peak

    can see a 360-degree panorama of the horizon."⏎"We really have an unobstructed view of the night sky. There's literally no light pollution

  4. #4 peak

    ophobia.⏎Photophobia is a painful over-sensitivity to light. Even the slightest amount of direct light

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. L3:2233 suppresses
  2. L3:2477 excites
  3. L4:2569 weakly suppresses
  4. L4:2498 weakly excites
  5. L4:1131 weakly excites
  6. L3:2654 weakly excites
  7. L2:1334 weakly suppresses
  8. L4:702 weakly suppresses
  9. L4:2522 weakly suppresses
  10. L4:2332 weakly excites
  11. L4:2698 weakly suppresses
  12. L4:2848 weakly suppresses
  13. L4:37 weakly suppresses
  14. L4:1298 weakly suppresses
  15. L3:1780 weakly suppresses
  16. L4:2703 weakly excites
  17. L2:1786 weakly suppresses
  18. L0:3039 weakly excites
  19. L4:229 weakly excites
  20. L0:545 weakly excites
  21. L0:2089 weakly suppresses
  22. L1:1352 weakly excites
  23. L3:2845 weakly suppresses
  24. L0:1995 weakly excites

Feeds later layers

  1. L20:3025 excites
  2. L8:1274 excites
  3. L12:2876 excites
  4. L19:1872 weakly suppresses
  5. L10:2444 weakly excites
  6. L21:1222 weakly excites
  7. L6:773 weakly suppresses
  8. L6:620 weakly suppresses "and" in lists of three or more
  9. L8:2495 weakly excites
  10. L10:1247 weakly excites
  11. L7:2518 weakly suppresses
  12. L11:2825 weakly suppresses
  13. L11:517 weakly excites
  14. L25:2895 weakly excites
  15. L23:239 weakly excites
  16. L9:2545 weakly suppresses
  17. L7:413 weakly excites
  18. L26:2223 weakly excites
  19. L15:2117 weakly excites
  20. L22:2448 weakly excites
  21. L24:2280 weakly excites
  22. L18:1255 weakly excites
  23. L10:1664 weakly excites
  24. L15:678 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 -1.25 and -1.121 values between -0.99 and -0.862 values between -0.86 and -0.721 values between -0.72 and -0.595 values between -0.59 and -0.4650 values between -0.46 and -0.33316 values between -0.33 and -0.191062 values between -0.19 and -0.061825 values between -0.06 and 0.07538 values between 0.07 and 0.2165 values between 0.21 and 0.3416 values between 0.34 and 0.476 values between 0.47 and 0.602 values between 0.60 and 0.736 values between 0.73 and 0.873 values between 0.87 and 1.002 values between 1.13 and 1.261 values between 1.26 and 1.401 values between 1.53 and 1.661 values between 1.66 and 1.791 values between 1.79 and 1.931 values between 2.06 and 2.191 values between 2.72 and 2.851 values between 3.91 and 4.05offnegativepositiveoften

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. lớp
  2. frente
  3. -layer
  4. 层
  5. 厚度
  6. layer
  7. меди
  8. orld
  9. 厚厚的
  10. 广告服务

Suppresses

  1. gles
  2. Desire
  3. �
  4. ean
  5. enas
  6. Rentals
  7. ansion
  8. 髑
  9. inals
  10. issions