Neuron Explorer

Qwen3-0.6B

blank space after a label colon

A neuron in layer 0 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 reacts to the empty spot right after a label and colon, like "Published:" or "Length:", where a value is about to appear. Turning it up did not clearly change the writing in a recognisable way.

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

    : 35-50 years⏎Length: 5 meters (15 feet

  2. #2 peak

    the moon.⏎- Ashura: 10th Muharram*⏎Ashura is an Arabic word meaning “

  3. #3 peak

    Number of native speakers: 13 million. …⏎- Arabic. Number of native speakers: 2

  4. #4 peak

    to draw out stubborn associations.<|endoftext|>Bacteria ‘evolving in the lab’?⏎‘A poke in the eye for anti-evolutionists’?⏎Published: 14 June 2008 (GMT+10)⏎Some bacteria

  5. #5 peak

    Next when you are done.⏎[Update: 3/21/2009] If you are interested in setting

  6. #6 peak

    0.3390/ijerph9114170⏎Published: 15 November 2012⏎Abstract: Understanding the effects of intensive

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: - I have been thinking about lately, and I have been thinking about lately. What is this something? So, this something is a type of something, and I have been thinking about lately. Now, the thing is a type of

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

    , might not be immediately obvious.⏎“It is not a part of our culture to think of

  2. #2 peak

    pets, might be the strongest natural material yet documented—even stronger than spider silk, the previous record

  3. #3 peak

    might think that this sounds like the theory of quarks. Yes and no — it

  4. #4 peak

    might also cause a serious crash with potential casualties.⏎You may recall the emergency landing of

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

None: this is the first layer.

Feeds later layers

  1. L3:77 excites
  2. L17:338 excites colon introducing an explanation
  3. L5:134 weakly excites
  4. L25:2082 weakly suppresses first-person 'I' starting a clause
  5. L19:1790 weakly excites
  6. L2:1470 weakly excites
  7. L2:1138 weakly excites
  8. L14:730 weakly suppresses
  9. L2:872 weakly excites
  10. L11:1391 weakly excites
  11. L19:1866 weakly suppresses
  12. L14:1544 weakly excites
  13. L18:296 weakly excites
  14. L12:501 weakly excites
  15. L12:553 weakly suppresses
  16. L20:624 weakly suppresses
  17. L17:2736 weakly excites
  18. L13:1705 weakly excites
  19. L27:1841 weakly excites
  20. L23:2902 weakly excites
  21. L12:265 weakly excites
  22. L22:3018 weakly excites
  23. L14:2431 weakly excites
  24. L19:2496 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 -0.33 and -0.311 values between -0.31 and -0.291 values between -0.27 and -0.253 values between -0.23 and -0.2110 values between -0.21 and -0.197 values between -0.19 and -0.1724 values between -0.17 and -0.1539 values between -0.15 and -0.1368 values between -0.13 and -0.11103 values between -0.11 and -0.09190 values between -0.09 and -0.07361 values between -0.07 and -0.05482 values between -0.05 and -0.03486 values between -0.03 and -0.01590 values between -0.01 and 0.01516 values between 0.01 and 0.03411 values between 0.03 and 0.05257 values between 0.05 and 0.07161 values between 0.07 and 0.0994 values between 0.09 and 0.1147 values between 0.11 and 0.1327 values between 0.13 and 0.1518 values between 0.15 and 0.176 values between 0.17 and 0.193 values between 0.19 and 0.211 values between 0.21 and 0.241 values between 0.46 and 0.48offnegativepositiveoften

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. orda
  2. []:⏎
  3. ctic
  4. ullan
  5. paged
  6. ucz
  7. ars
  8. qua
  9. /lgpl
  10. уч

Suppresses

  1. TestingModule
  2. 然是
  3. 投放
  4. 벅
  5. 之人
  6. Uploaded
  7. 在家里
  8. lude
  9. Endpoints
  10. endpoints