Neuron Explorer

Qwen3-0.6B

document break token mid-sentence

A neuron in layer 21 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 most strongly to the invisible marker that separates two unrelated documents, especially when it lands in the middle of a sentence. Its negative side reacts to 'i.e' abbreviations. Turning it up or down mostly made the text degrade into repeated punctuation or unrelated talk, so the steering shows no 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

    of workers to a consultancy. These consultancies offer comprehensive services, which normally include –⏎• Activities related to registration and licensing necessary under the labour laws,⏎•<|endoftext|>- Making Maths Accessible⏎Version Dated 24th February 2

  2. #2 peak

    rounded to the nearest integer.<|endoftext|>Christine Duncan, MD⏎Sickle cell disease (SCD) affects as

  3. #3 peak

    a phase boundary, for instance, gas-oil, oil-water, and<|endoftext|>Green kitchen: A recipe for being more environmentally friendly⏎Cooks can do their

  4. #4 peak

    they eat, so to avoid causing spikes in their blood glucose, but type 1 cannot be controlled solely with diet. The necessity for treatment with insulin is why<|endoftext|>Last updated: July 27, 2007July 2

  5. #5 peak

    chool.org/), our experiment with graphics quickly turned into a full-grown graphic anthropology workshop, and some of the participants used sketching in ways that did not<|endoftext|>People vary in their reactions to mosquito bites. Most people develop itchy, raised

  6. #6 peak

    . It is because the re-exported goods might require some additional information.⏎Having understood so much about reexport trade and what are re-exportation, it is<|endoftext|>10.2.2 Restarting TCP/IP⏎TCP/IP is a peer

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:.,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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

    is a two-letter rating that tells you the type of engine the oil is designed for (i.e. gasoline or diesel) and the quality level. The “S” designation (

  2. #2 peak

    type 1 and also non insulin dependent diabetes or diabetes type 2.⏎Diabetes is worldwide in distribution and the incidence of both types of primary diabetes, i.e insulin dependent diabetes mellitus (IDDM) or diabetes mellitus type 1

  3. #3 peak

    . For example, if the array is [1, 2, 3, 4, 5], it will print it in reverse order i.e. from the last element to the first element:⏎5 4 3

  4. #4 peak

    of selective toxicity: that certain chemicals that would be toxic to some organisms, e.g., infectious bacteria, would be harmless to other organisms, e.g., humans. In 1928, Sir Alexander Fleming, a Scottish

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. L17:152 strongly excites mid-word fragments, often 'intric'
  2. L17:70 excites
  3. L6:293 suppresses
  4. L16:1421 suppresses
  5. L19:696 excites
  6. L10:43 excites "terms" in "in terms of"
  7. L12:50 suppresses
  8. L4:1572 excites
  9. L11:42 suppresses
  10. L7:9 suppresses
  11. L18:91 excites
  12. L14:102 excites
  13. L19:220 suppresses question mark ending a web question heading
  14. L5:2525 suppresses
  15. L8:62 suppresses
  16. L19:1708 excites
  17. L19:377 suppresses
  18. L20:131 suppresses
  19. L14:56 suppresses document ends and list line breaks
  20. L1:78 excites
  21. L5:34 excites
  22. L20:472 excites
  23. L6:1031 excites
  24. L5:1127 excites

Feeds later layers

  1. L27:156 suppresses
  2. L24:94 suppresses
  3. L25:30 suppresses
  4. L24:385 suppresses
  5. L24:39 suppresses
  6. L26:356 suppresses
  7. L27:151 suppresses
  8. L26:21 excites
  9. L22:765 suppresses
  10. L23:191 suppresses
  11. L22:14 suppresses
  12. L23:676 suppresses the 'm' in a.m. and p.m
  13. L22:424 suppresses
  14. L22:65 suppresses
  15. L24:34 excites
  16. L24:31 excites
  17. L24:508 suppresses
  18. L25:17 suppresses
  19. L23:80 suppresses blank slot before a year or number
  20. L27:11 excites unclear: mixed fragments in lists and pronunciations
  21. L27:301 suppresses
  22. L27:342 suppresses
  23. L26:131 excites
  24. L22:1733 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.

2 values between -15.91 and -14.972 values between -14.03 and -13.094 values between -13.09 and -12.157 values between -12.15 and -11.214 values between -11.21 and -10.273 values between -10.27 and -9.3311 values between -9.33 and -8.3813 values between -8.38 and -7.4415 values between -7.44 and -6.5018 values between -6.50 and -5.5617 values between -5.56 and -4.6231 values between -4.62 and -3.6828 values between -3.68 and -2.7434 values between -2.74 and -1.8067 values between -1.80 and -0.86166 values between -0.86 and 0.091784 values between 0.09 and 1.031624 values between 1.03 and 1.9739 values between 1.97 and 2.9115 values between 2.91 and 3.855 values between 3.85 and 4.793 values between 4.79 and 5.733 values between 5.73 and 6.682 values between 6.68 and 7.621 values between 7.62 and 8.561 values between 9.50 and 10.442 values between 10.44 and 11.381 values between 12.32 and 13.261 values between 16.09 and 17.031 values between 17.03 and 17.971 values between 17.97 and 18.911 values between 18.91 and 19.851 values between 19.85 and 20.791 values between 20.79 and 21.73offnegativepositiveoften

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. ,
  2. �
  3. not
  4. `
  5. **
  6. \
  7. se
  8. IA
  9. **
  10. and

Suppresses

  1. @update
  2. uai
  3. ableObject
  4. 子里
  5. ',...⏎
  6. etting
  7. ']->
  8. usses
  9. 了他的
  10. ",-