Neuron Explorer

Qwen3-0.6B

sentence end before a new paragraph

A neuron in layer 18 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 the full stop at the end of an introductory sentence in an article, the point where the next paragraph is about to start. Turning it up or down gave repetitive or rambling text, so the effect on writing 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

    of obsidian). However, sometimes obsidian isn’t as readily identifiable as you might think. It can grade into other rock types, making it difficult to distinguish.⏎Perhaps you have found a less common variety that doesn’t match your preconceived

  2. #2 peak

    10 appearing for the board exams. Students can easily NCERT Solutions for Class 10 Maths Exercise 6.5 Chapter 6 Triangles from below.⏎CBSE Class 10 Maths Chapter 6 Triangles Exercise 6.

  3. #3 peak

    that causes immune system dysfunction and multi-organ failure. Alverdy had a hunch that gut microbes might play a critical and culpable role in this process.⏎Fifteen years ago, Alverdy teamed up with Olga Zabor

  4. #4 peak

    this blog, we’ll discuss all the important aspects of wearing shockproof gloves, starting from electric hazards, requirements for the right PPE, and tips and tricks.⏎Common Electrical Hazards⏎To discuss common electrical hazards that require the use of shock

  5. #5 peak

    have an aneuploidy. Recent recommendations from the American College of Obstetricians and Gynecologists (ACOG) are quite clear on that issue.⏎In those same recommendations, ACOG also states

  6. #6 peak

    care of them until they are large enough to transplant out.⏎It’s really not that difficult. Once you have the supplies and some minimal tools, you just start.⏎So let’s have a look at what you need and how to get started.⏎

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: of the only possible remaining use after all else, not at all, and/o/o, rather,, rather, rather,/o/o,,, (, ,, ,, ,, ,, ,, ,, ,, ,

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

    |⏎|Order Information:||Web: http://www.tandfonline.com/pricing/journal/VECE20|⏎When requesting a correction, please mention this item's handle: RePEc:taf:jeduce:v:

  2. #2 peak

    , access via your institution.⏎Buy single article⏎Instant access to the full article PDF.⏎Price includes VAT (USA)⏎Tax calculation will be final

  3. #3 peak

    is an extension of Bullseye matrix problem.⏎Who has power to do everything in this world?⏎Close MATLAB with keyboard⏎Find the treasures in MATLAB Central and discover how the community can help you!⏎Choose a web site to get translated content where

  4. #4 peak

    status as electors in August 1920 by the Nineteenth Amendment to the federal Constitution, the...⏎Users without a subscription are not able to see the full⏎to access all content.<|endoftext|>Great Slaty becomes World's

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. L9:25 excites
  2. L14:1142 excites
  3. L8:32 excites
  4. L11:29 excites
  5. L15:1272 suppresses
  6. L7:22 weakly suppresses
  7. L13:414 weakly excites
  8. L5:726 weakly excites
  9. L11:864 weakly excites
  10. L13:2946 weakly suppresses
  11. L8:304 weakly suppresses
  12. L10:61 weakly suppresses
  13. L5:828 weakly excites
  14. L13:2712 weakly excites
  15. L9:2447 weakly excites
  16. L5:986 weakly suppresses
  17. L17:892 weakly suppresses
  18. L15:81 weakly suppresses line breaks inside running text
  19. L14:1800 weakly excites
  20. L17:794 weakly excites
  21. L6:3051 weakly suppresses
  22. L10:259 weakly excites
  23. L9:1118 weakly excites
  24. L15:1406 weakly suppresses

Feeds later layers

  1. L22:67 excites
  2. L26:114 excites
  3. L19:1619 suppresses
  4. L27:22 excites
  5. L21:100 excites
  6. L22:282 weakly excites period then line break in lists
  7. L21:524 weakly excites
  8. L23:407 weakly excites
  9. L23:179 weakly excites
  10. L21:1854 weakly suppresses
  11. L22:506 weakly excites
  12. L19:2709 weakly excites
  13. L22:418 weakly excites
  14. L24:261 weakly excites
  15. L22:1242 weakly excites
  16. L21:113 weakly excites
  17. L19:155 weakly excites
  18. L23:262 weakly excites
  19. L19:601 weakly suppresses
  20. L27:661 weakly excites
  21. L27:65 weakly excites
  22. L19:508 weakly excites
  23. L26:224 weakly excites
  24. L19:2404 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.

3 values between -5.36 and -4.446 values between -3.52 and -2.609 values between -2.60 and -1.6833 values between -1.68 and -0.762579 values between -0.76 and 0.171038 values between 0.17 and 1.0939 values between 1.09 and 2.0124 values between 2.01 and 2.9312 values between 2.93 and 3.8515 values between 3.85 and 4.7711 values between 4.77 and 5.706 values between 5.70 and 6.625 values between 6.62 and 7.545 values between 7.54 and 8.4616 values between 8.46 and 9.3810 values between 9.38 and 10.308 values between 10.30 and 11.2320 values between 11.23 and 12.1513 values between 12.15 and 13.0710 values between 13.07 and 13.995 values between 13.99 and 14.915 values between 14.91 and 15.832 values between 15.83 and 16.763 values between 16.76 and 17.686 values between 17.68 and 18.606 values between 18.60 and 19.526 values between 19.52 and 20.442 values between 20.44 and 21.361 values between 21.36 and 22.281 values between 22.28 and 23.211 values between 23.21 and 24.133 values between 24.13 and 25.052 values between 25.97 and 26.892 values between 26.89 and 27.811 values between 30.58 and 31.50offnegativepositiveoften

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. Indeed
  2. Figure
  3. 我们可以
  4. 这就是
  5. 至於
  6. 有一种
  7. According
  8. Indeed
  9. 不仅如此
  10. Fig

Suppresses

  1. itional
  2. ereum
  3. �始
  4. OMIC
  5. :variables
  6. omedical
  7. iac
  8. IColor
  9. 색
  10. thic