Neuron Explorer

Qwen3-0.6B

line break after a section heading

A neuron in layer 16 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

The neuron fires on the line break that comes right after a heading or label, such as 'ASJC Scopus subject areas' or 'Similar books and articles', just before a list starts. Turning it up made the model write in short broken lines, so the effect is hard to read.

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

    coated and graded-index-layer coated samples was performed and compared with experimentally derived films. Good agreement between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Ceramics and Composites⏎-

  2. #2 peak

    silkworm-BEVS is a promising alternative strategy to produce the functional and thermostable reverse transcriptase.⏎All Science Journal Classification (ASJC) codes⏎- Insect Science<|endoftext|>An account of the religion of

  3. #3 peak

    Seeking Help:⏎Ben Locke, Ph.D., a mental health expert serving at Pennsylvania State University,

  4. #4 peak

    Muscle Growth⏎One of the⏎key aspects of a successful exercise program for muscle growth is exercise

  5. #5 peak

    12|⏎|Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most computer users are familiar with the

  6. #6 peak

    and Surveys⏎- NARA Energy Literacy Principles Matrix (a bioenergy resource search tool for

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

    first⏎> place. If one is chasing through a data structure and finds a need to⏎> change something buried within it, it seems necessary to rebuild⏎> everything that includes the changed thing.⏎In general, values are immutable

  2. #2 peak

    ⏎soluble vitamins obtainable no other way, and are even part of cell wall⏎structure. Oils and Fats can also have very negative effects on⏎Before the first permanent settlements, nutrient oils came from insects,

  3. #3 peak

    also describes key elements of the Java API⏎library, such as I/O, the Collections Framework, the concurrency⏎utilities, the stream library, networking, event⏎handling, AWT, and more. Three chapters provide a solid introduction⏎to

  4. #4 peak

    the Billund⏎headquarters, where the company employs approximately 120 designers.⏎The company also has smaller design offices in the UK, Spain, Germany,⏎and Japan, which are tasked with developing products aimed⏎specifically at these markets

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. L15:81 suppresses line breaks inside running text
  2. L10:230 suppresses
  3. L12:1012 excites
  4. L7:239 excites
  5. L7:2925 suppresses
  6. L8:88 excites line breaks inside wrapped sentences
  7. L8:2432 suppresses line breaks in lists and wrapped text
  8. L10:1760 suppresses
  9. L10:2315 excites line break inside wrapped text
  10. L14:56 suppresses document ends and list line breaks
  11. L1:1279 suppresses
  12. L6:463 excites
  13. L9:234 suppresses line break inside hard-wrapped text
  14. L8:380 weakly suppresses
  15. L9:1515 weakly excites line break splitting a phrase
  16. L14:51 weakly excites colon that introduces a statement
  17. L14:191 weakly excites blank spot before a run-on number
  18. L2:489 weakly suppresses
  19. L7:2952 weakly excites line break mid-sentence
  20. L8:14 weakly excites
  21. L10:84 weakly suppresses
  22. L13:10 weakly excites
  23. L13:309 weakly excites
  24. L1:1340 weakly suppresses

Feeds later layers

  1. L17:1242 strongly suppresses
  2. L17:1564 suppresses
  3. L18:1996 suppresses
  4. L19:970 suppresses
  5. L20:2399 suppresses
  6. L19:901 excites
  7. L26:1309 excites
  8. L21:579 excites
  9. L18:795 excites
  10. L19:771 excites
  11. L24:1619 excites
  12. L17:1675 weakly suppresses
  13. L25:2204 weakly excites
  14. L17:2160 weakly excites
  15. L26:664 weakly excites
  16. L23:1723 weakly excites
  17. L27:1467 weakly excites
  18. L24:203 weakly excites
  19. L27:213 weakly excites
  20. L17:2285 weakly excites
  21. L27:860 weakly excites
  22. L25:1805 weakly excites
  23. L17:825 weakly excites
  24. L27:106 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.

1 values between -12.28 and -11.911 values between -8.52 and -8.151 values between -5.52 and -5.142 values between -3.64 and -3.271 values between -3.27 and -2.891 values between -2.51 and -2.141 values between -2.14 and -1.762 values between -1.76 and -1.393 values between -1.39 and -1.018 values between -1.01 and -0.6430 values between -0.64 and -0.261988 values between -0.26 and 0.121776 values between 0.12 and 0.4991 values between 0.49 and 0.871 values between 0.87 and 1.241 values between 2.37 and 2.75offnegativepositiveoften

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. DN
  2. isbn
  3. 纶
  4. DN
  5. _Do
  6. dn
  7. ^⏎
  8. sworth
  9. dư
  10. ENDED

Suppresses

  1. hä
  2. ScrollIndicator
  3. pher
  4. rections
  5. createQuery
  6. oples
  7. quired
  8. idd
  9. createCommand
  10. 偏偏