Neuron Explorer

Qwen3-0.6B

period ending a sentence in essays

A neuron in layer 15 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 full stop at the end of a sentence in ordinary web prose, especially where a new point or claim is about to begin. Turning it up made the model write in odd, off-topic directions, and turning it down made it repeat itself.

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

    interpreting Scripture. Course Objectives: 1. The student will demonstrate an understanding of the basic rules and principles of Biblical hermeneutics. 2.⏎Additional Physical Format: Online version: Berkhof, Louis, Principles of Biblical

  2. #2 peak

    had absolutely nothing worry but fear itself” (pg. 6) c) “absolutely nothing to purchase and no money to purchase with it” (pg.⏎6) d) “nothing to see outside the boundaries of Maycomb County”

  3. #3 peak

    . The patients who survived were often quiet, but generally this surgery made the patients worse. Today lobotomy is considered barbaric and it is not practiced in Norway.⏎From a window⏎Lier Psyciatric Hospital, or Lier Asylum

  4. #4 peak

    and experience life. But there is good news. Depression is highly treatable. Studies show that a combination of antidepressant medication and psychotherapy is the best prescription.⏎One very common misconception about depression is that it feels like mere sadness. Sadness

  5. #5 peak

    Bottled water is a safe, healthy, and convenient packaged food product, which is comprehensively regulated at both the federal and state level.⏎Strictly regulated as a packaged food product by the U.S. Food and Drug

  6. #6 peak

    is the Tata Employees Union (TEU) formed in 1935. There are two ranks of the workforce at Tata Steel, i.⏎e., officers and non-officers (workers who work in mines and plants and

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 using this pattern and I have been working on this pattern and I have been working on and I have been working on, and I have been working on and I have been working on and I have been working on and 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

    the field from August 2 – August 20.<|endoftext|>Genetics as it applies to evolution, molecular biology, and medical aspects.⏎10 posts • Page 1 of 1⏎Hope somebody can help, I've got an A

  2. #2 peak

    in which it appears.”⏎For more information, The report card can be found online: www.nationsreportcard.gov.MORE IN Vermont News⏎- Most Popular⏎- Most Emailed⏎- MEDIA GALLERY<|endoftext|>Chim's striking image

  3. #3 peak

    (6-8) Show all Grade Level⏎Resource Type: Activities⏎Results 1 - 7 of 7 matches⏎From Grid to Home⏎Lisa Gardiner, National Center for Atmospheric Research; Marie

  4. #4 peak

    - Filed Under⏎J ennifer Skaja has seen the signs of small miracles, as children with

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. L7:45 suppresses
  2. L14:1256 suppresses
  3. L4:2258 suppresses
  4. L14:56 suppresses document ends and list line breaks
  5. L4:2219 excites
  6. L6:2679 suppresses
  7. L10:157 suppresses end of document and web boilerplate
  8. L6:2905 weakly suppresses
  9. L10:2887 weakly excites
  10. L4:703 weakly suppresses
  11. L13:10 weakly excites
  12. L10:1578 weakly excites
  13. L9:1167 weakly suppresses
  14. L9:1301 weakly excites
  15. L9:1528 weakly suppresses
  16. L12:771 weakly excites
  17. L5:2544 weakly suppresses
  18. L4:2337 weakly suppresses
  19. L14:1411 weakly excites
  20. L5:780 weakly suppresses
  21. L9:7 weakly excites
  22. L0:1449 weakly suppresses
  23. L14:2288 weakly excites
  24. L7:58 weakly suppresses

Feeds later layers

  1. L18:1341 excites
  2. L16:944 excites
  3. L17:2317 excites
  4. L19:301 excites
  5. L16:592 excites
  6. L23:374 excites
  7. L19:2060 excites
  8. L16:464 weakly excites
  9. L16:1371 weakly suppresses
  10. L16:190 weakly excites end-of-document boundary token
  11. L19:1940 weakly excites
  12. L21:2057 weakly excites
  13. L23:837 weakly suppresses
  14. L17:2674 weakly excites
  15. L19:2031 weakly excites line breaks in web footers and metadata
  16. L16:2046 weakly excites
  17. L17:674 weakly excites
  18. L16:1204 weakly excites
  19. L18:1689 weakly excites
  20. L17:436 weakly excites
  21. L17:849 weakly excites
  22. L18:2486 weakly excites
  23. L16:91 weakly excites
  24. L26:2952 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.

9 values between -1.28 and -0.96100 values between -0.96 and -0.64978 values between -0.64 and -0.322231 values between -0.32 and -0.00367 values between -0.00 and 0.3248 values between 0.32 and 0.6321 values between 0.63 and 0.9530 values between 0.95 and 1.2716 values between 1.27 and 1.5917 values between 1.59 and 1.9113 values between 1.91 and 2.2311 values between 2.23 and 2.545 values between 2.54 and 2.864 values between 2.86 and 3.189 values between 3.18 and 3.509 values between 3.50 and 3.825 values between 3.82 and 4.134 values between 4.13 and 4.456 values between 4.45 and 4.772 values between 5.09 and 5.411 values between 5.41 and 5.734 values between 5.73 and 6.042 values between 6.04 and 6.364 values between 6.36 and 6.685 values between 6.68 and 7.004 values between 7.00 and 7.321 values between 7.32 and 7.641 values between 8.27 and 8.591 values between 11.14 and 11.45offnegativepositiveoften

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. ponde
  2. oh
  3. osi
  4. ion
  5. هل
  6. occo
  7. interest
  8. hab
  9. их
  10. よい

Suppresses

  1. (Cs
  2. \Input
  3. whites
  4. älle
  5. \Repositories
  6. ptive
  7. ringe
  8. PointerException
  9. undler
  10. leneck