Neuron Explorer

Qwen3-0.6B

line break in the middle of a sentence

A neuron in layer 2 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 a line break that cuts a sentence in two, the kind you see in text that was wrapped at a fixed width (like old printed articles). It is most active where the sentence is clearly unfinished at that point. Turning it up gave text that does not obviously relate to this.

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

    policy shifts in member states,⏎such as Greece, Italy the United Kingdom and Ireland. The Multireader⏎project examined complex navigation through digital materials, and the⏎TEDUB project covered a similar area for graphical information

  2. #2 peak

    Site 48TE414 is in the South Jenny Lake development. The site was⏎recommended as eligible for the National Register of Historic Places. The⏎National Park Service and

  3. #3 peak

    and heavens shall be replaced by a new earth and new heavens; a new⏎world will then come into being. All mankind shall be raised up there and shall⏎be rewarded or punished on the basis of their beliefs and deeds.”⏎In this article

  4. #4 peak

    .⏎The ceiling limit for a family consisting of five members was fixed as 30⏎standard acres. For every additional member of the family consisting of more than⏎five members, an additional extent of

  5. #5 peak

    99⏎African American Nurses have long history with the American Red -⏎Undeterred by initial rejection, Davis persisted and in l9l7 became the⏎first

  6. #6 peak

    plantation after getting permission from the Land⏎Board under section 31 of the Act.⏎Ceiling laws were enacted and enforced actually in two phases, the earlier⏎

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: $$ \begin{array}{l} \sum_{i} = \sum_{i} $$\sum_{i} $$\sum_{i} $$\sum_{i} $$ $$ \

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

    which ids full of⏎violence and hatred, wreaking havoc on their towns to express their beliefs.The Oklahoma City bombing brought home to American people that no country⏎is immune from terrorism.⏎Although thought to be tied into larger extremist⏎groups,

  2. #2 peak

    the⏎notes are a summary and a detailed analysis of the⏎poem's plot, theme, genre, purpose, context, references and techniques. The original⏎text appears after the notes. For "deep explorers" of Anglo-Saxon

  3. #3 peak

    <|endoftext|>Lecture by Francesca Russello Ammon, Associate Professor of City & Regional Planning and Historic Preservation, Weitzman School of Design, University of Pennsylvania⏎By 1965, nearly

  4. #4 peak

    sharing, and rapidly publishing scholarly content on the Web.” It is not simply a Web site that hosts textbooks, articles, and courses. A non-profit run by⏎“Most textbooks are a mass of information in linear format: one topic follows after

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. L1:167 excites
  2. L1:1143 weakly excites
  3. L0:107 weakly suppresses newline starting a new block
  4. L0:145 weakly suppresses
  5. L1:292 weakly excites
  6. L0:1199 weakly excites
  7. L1:685 weakly excites
  8. L1:2862 weakly suppresses
  9. L1:2228 weakly excites
  10. L0:2669 weakly suppresses
  11. L0:1510 weakly suppresses
  12. L1:1278 weakly suppresses
  13. L0:3042 weakly suppresses
  14. L1:863 weakly excites
  15. L1:203 weakly suppresses
  16. L1:2206 weakly suppresses
  17. L1:1960 weakly excites
  18. L1:2313 weakly suppresses
  19. L0:1880 weakly excites
  20. L1:1279 weakly suppresses
  21. L1:2564 weakly excites
  22. L1:2724 weakly suppresses
  23. L1:1737 weakly excites
  24. L0:224 weakly excites

Feeds later layers

  1. L3:1901 excites
  2. L8:2674 weakly suppresses
  3. L24:1683 weakly suppresses
  4. L17:242 weakly excites
  5. L8:380 weakly excites
  6. L22:2867 weakly suppresses
  7. L9:589 weakly suppresses
  8. L7:1665 weakly suppresses
  9. L27:365 weakly suppresses
  10. L3:1454 weakly suppresses
  11. L10:1399 weakly excites
  12. L10:1086 weakly suppresses
  13. L4:1749 weakly suppresses
  14. L22:167 weakly excites
  15. L8:896 weakly suppresses
  16. L3:1036 weakly excites
  17. L3:1592 weakly excites
  18. L5:342 weakly excites line breaks inside wrapped sentences
  19. L23:135 weakly excites
  20. L9:2361 weakly suppresses
  21. L15:939 weakly suppresses
  22. L12:911 weakly suppresses
  23. L24:935 weakly suppresses
  24. L11:925 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.

1 values between -1.40 and -1.331 values between -1.20 and -1.131 values between -0.86 and -0.791 values between -0.72 and -0.662 values between -0.66 and -0.591 values between -0.59 and -0.521 values between -0.52 and -0.453 values between -0.45 and -0.3811 values between -0.38 and -0.3234 values between -0.32 and -0.25110 values between -0.25 and -0.18245 values between -0.18 and -0.11572 values between -0.11 and -0.051065 values between -0.05 and 0.02870 values between 0.02 and 0.09565 values between 0.09 and 0.16285 values between 0.16 and 0.23100 values between 0.23 and 0.2926 values between 0.29 and 0.3610 values between 0.36 and 0.431 values between 0.56 and 0.631 values between 0.63 and 0.702 values between 1.24 and 1.31offnegativepositiveoften

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. &,
  2. distilled
  3. lectic
  4. comings
  5. kest
  6. вели
  7. 过敏
  8. utas
  9. учет
  10. łość

Suppresses

  1. واب
  2. ibre
  3. wings
  4. اث
  5. Directorate
  6. dav
  7. Titanium
  8. athed
  9. ье
  10. Sheets