Neuron Explorer

Qwen3-0.6B

line break splitting a phrase

A neuron in layer 9 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 lights up on a line break that cuts a sentence or name in the middle, where the words after the break finish the same thought. Steering did not reveal a clear theme, so the link to the output is unclear.

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

    e⏎Southern black tit male, Thornybush Game Reserve,⏎South Africa. [photo Trevor Hardaker ©]⏎Southern black tit female, Thornybush Game⏎Reserve, South Africa. [photo Trevor Hardaker ©]⏎Distribution and habitat

  2. #2 peak

    ⏎the seed at the appropriate time and rate⏎for your species.⏎to spread seed⏎- Measure the area to be planted and determine the total amount of seed⏎-

  3. #3 peak

    the problem⏎Each member of the team takes a card from the pile in turn (one⏎card each) without the others seeing it.⏎If you have an⏎observer, they might find the following list of things to look for⏎helpful

  4. #4 peak

    from 'The Photographic Society'⏎in 1892 because of it's lack of interest in art photography. The first show of The Linked Ring Brotherhood⏎in November 1893 was called the Photographic Salon⏎and included

  5. #5 peak

    DIFFERENCES IN THESE COMMON SLEEP ISSUES⏎Snoring is often one of the first indicators of sleep apnea, but not everyone who snores has⏎What is snoring?⏎Snoring occurs when your throat muscles relax, allowing the

  6. #6 peak

    and improving the healthcare of⏎Nursing Old Wounds - Recollections⏎by Nancy Leftenant-Colon, as the first black nurse in the Regular Nurse⏎Corps in 1942.⏎The Crile Archives Current Project -⏎

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: The key is to focus on the first 10. I think the problem is that the first 10 should focus on the first 10, but the problem is the first 10, the problem is the first

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

    extra step is required to terminate the for loop,⏎> hence n + 1 and not n executions), which will consume T4( n + 1 )⏎> time." (from http://en.wikipedia.org/wiki/Analysis_of_algorithms)⏎

  2. #2 peak

    the Patent Counsel, Johnson Space Center, (281) 483-0837. Refer to MSC-23

  3. #3 peak

    <|endoftext|>Let’s Talk Fatty Liver Disease⏎When we think about essential bodily organs, there are two primary candidates: our brain and our heart. However, every single organ in the body has a role in something, which makes every organ essential.

  4. #4 peak

    15⏎Isothermal Community College⏎P.O. Box 804, Spindale NC 28160⏎(828) 395-1307 fax (828)

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. L8:982 suppresses
  2. L7:223 excites
  3. L7:2925 suppresses
  4. L8:1468 excites
  5. L8:926 excites
  6. L6:1540 excites
  7. L6:181 excites periods in abbreviations and the 's in names
  8. L8:1396 excites
  9. L6:199 weakly excites
  10. L7:355 weakly suppresses
  11. L7:173 weakly excites
  12. L7:948 weakly excites
  13. L6:43 weakly suppresses
  14. L6:154 weakly excites
  15. L8:419 weakly suppresses
  16. L8:1208 weakly suppresses
  17. L3:439 weakly suppresses
  18. L7:1354 weakly suppresses
  19. L4:212 weakly excites
  20. L7:1771 weakly suppresses
  21. L8:797 weakly excites
  22. L8:3048 weakly suppresses
  23. L7:616 weakly excites
  24. L5:1962 weakly excites

Feeds later layers

  1. L12:792 excites
  2. L13:1108 excites
  3. L16:2042 excites
  4. L15:2449 excites line breaks mid-sentence
  5. L21:348 excites mid-sentence line breaks
  6. L15:638 excites "the" inside institution names
  7. L12:1112 excites
  8. L14:493 excites
  9. L17:310 excites
  10. L10:152 excites
  11. L18:617 excites
  12. L15:3057 excites
  13. L10:404 excites
  14. L12:1823 excites
  15. L15:646 excites
  16. L18:933 excites
  17. L15:2046 excites
  18. L15:168 excites
  19. L20:526 excites
  20. L16:2076 excites
  21. L15:1846 excites
  22. L15:2021 excites
  23. L20:90 excites
  24. L15:998 excites period inside an abbreviation

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 -1.06 and -0.932 values between -0.93 and -0.804 values between -0.80 and -0.674 values between -0.67 and -0.5419 values between -0.54 and -0.4183 values between -0.41 and -0.28371 values between -0.28 and -0.151064 values between -0.15 and -0.021508 values between -0.02 and 0.11666 values between 0.11 and 0.24141 values between 0.24 and 0.3724 values between 0.37 and 0.507 values between 0.50 and 0.632 values between 0.63 and 0.762 values between 0.76 and 0.891 values between 0.89 and 1.021 values between 1.02 and 1.151 values between 1.28 and 1.412 values between 1.54 and 1.671 values between 2.31 and 2.441 values between 2.44 and 2.571 values between 2.57 and 2.701 values between 4.00 and 4.13offnegativepositiveoften

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. ún
  2. ạch
  3. EITHER
  4. чь
  5. bölgesinde
  6. ecessarily
  7. AdminController
  8. ước
  9. ChangedEventArgs
  10. Ẹ

Suppresses

  1. ⏎
  2. ctic
  3. iga
  4. iae
  5. iative
  6. _
  7. <p
  8. 徒
  9. iosa