Neuron Explorer

Qwen3-0.6B

"and" joining paired nouns

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 fires on the word "and" when it links two nouns in a pair, such as "signs and symptoms" or "vitamins and minerals", especially in formal or technical writing. Turning it up did not produce a clear, matching change in the text.

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

    Teachers, a resource providing teachers with guidance on how to effectively incorporate YouTube and video into their classrooms. “As a company committed to the next generation of scientists

  2. #2 peak

    /1/77065/10 by the strategic scientific research and experimental development program:⏎SYNAT - “Interdisciplinary System for Interactive Scientific and Scientific-Technical Information”.<|endoftext|>An team of researchers heading to the Antarctic later this

  3. #3 peak

    “While state laws and educational programs are critical, ultimately, parents are the most critical component to keep their

  4. #4 peak

    Registry.⏎—Three definitions of lifetime prevalence of alcoholism based on Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised criteria: (1) alcohol

  5. #5 peak

    liver damage has occurred.⏎These signs and symptoms are features of liver cirrhosis ("hardening of the liver tissue")

  6. #6 peak

    multivitamin can help to work towards your regular daily requirements for essential vitamins and minerals, if dietary intake is inadequate.⏎What Can Vitamins Do for Your Health

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... about it. So I have been thinking about... So I have been thinking about... So I have been thinking about... So I have been thinking about... So I have

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

    , though a death camp like Westerbork, one that also<|endoftext|>Dangerously Sleepy explores the fraught relations between overwork, sleep deprivation, and public health. Health and labor historian Alan Derickson charts the cultural and political

  2. #2 peak

    motivating change and encouraging pro-environmental action;⏎Enhance communication between experts, legislators, and lay audiences concerning environmental issues;⏎Inspire educators in improving environmental education curricula;⏎

  3. #3 peak

    AgreementArticle Free Pass⏎Constantinople Agreement, (March 18, 1915), secret World War I agreement between Russia, Britain, and France for the

  4. #4 peak

    actual picture of the Biggest Beaver Dam in the World:<|endoftext|>Editorial and Revision Process⏎Preparing an Inquiry article for publication is a collaboration between authors, editors, and faculty mentors. The authors of Inquiry articles often begin with a report or essay describing

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. L18:870 excites
  2. L17:572 excites
  3. L21:735 excites 'or' in ranges and alternatives
  4. L16:376 excites 'or' joining alternatives
  5. L9:92 excites "and" or "or" joining two items
  6. L15:2765 excites
  7. L19:2711 excites
  8. L19:2080 excites ampersand joining two items
  9. L3:433 excites
  10. L20:885 excites 'and' joining paired words
  11. L13:131 excites 'and' joining parallel phrases
  12. L1:1566 suppresses
  13. L16:111 excites
  14. L26:1007 excites
  15. L17:2274 excites
  16. L17:1537 excites
  17. L3:347 excites 'and' joining two words
  18. L12:1786 excites
  19. L16:2049 weakly excites
  20. L8:70 weakly excites 'and' between stacked superlatives
  21. L11:406 weakly excites
  22. L22:198 weakly excites
  23. L4:131 weakly excites
  24. L9:163 weakly excites

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.

8 values between -0.64 and -0.525 values between -0.52 and -0.408 values between -0.40 and -0.2822 values between -0.28 and -0.16597 values between -0.16 and -0.042904 values between -0.04 and 0.08256 values between 0.08 and 0.2012 values between 0.20 and 0.321 values between 0.44 and 0.561 values between 0.56 and 0.681 values between 0.68 and 0.801 values between 0.80 and 0.922 values between 1.04 and 1.162 values between 1.29 and 1.412 values between 1.53 and 1.653 values between 1.65 and 1.774 values between 1.77 and 1.893 values between 1.89 and 2.012 values between 2.01 and 2.138 values between 2.13 and 2.253 values between 2.25 and 2.376 values between 2.37 and 2.497 values between 2.49 and 2.617 values between 2.61 and 2.7310 values between 2.73 and 2.858 values between 2.85 and 2.976 values between 2.97 and 3.094 values between 3.09 and 3.214 values between 3.21 and 3.334 values between 3.33 and 3.454 values between 3.45 and 3.571 values between 3.57 and 3.691 values between 3.81 and 3.931 values between 4.05 and 4.17offnegativepositiveoften

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. /or
  2. equally
  3. files
  4. 洽
  5. mutually
  6. дей
  7. kok
  8. ur
  9. phans
  10. gt

Suppresses

  1. anno
  2. ода
  3. apixel
  4. 钡
  5. Ital
  6. .").
  7. entifier
  8. 玳
  9. ;");⏎
  10. Wolver