Neuron Explorer

Qwen3-0.6B

line breaks inside running text

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 reacts to line breaks that fall in the middle of a sentence or list item, where the text carries on after the break. It was strongest on those breaks, and it went quiet, even negative, on breaks that come right after a question mark ending a list line. Turning it up made the model fall into long runs of blank lines, but the steered text was otherwise 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

    off Victory|⏎Federal Elections (94-2004): Election Results for the US Senate⏎and the US House of Representatives. Copyright of the Federal⏎Elections

  2. #2 peak

    early through its center; by Mad and White Rivers; and by Lake Champlain,⏎which affords it many navigable privileges. The soil is good, particularly⏎in those towns below the mountains, and bordering the lake<|endoftext|>“ The L

  3. #3 peak

    can⏎be fitted together?⏎Using different numbers of sticks, how many different triangles are⏎you able to make? Can you make any rules about the numbers of⏎sticks that make the most triangles?⏎Design an arrangement of display boards in

  4. #4 peak

    the Kupa Piti Kungka Tjuta, the Senior Aboriginal⏎Women from Coober Pedy, South Australia. Documents their knowledge of atomic⏎tests in the 1950's and 60's and their

  5. #5 peak

    things to look for⏎helpful and feed back to the team at the end of the task.⏎- Team members cannot show their cards to anyone else on the⏎- A team member can describe what is on their card.⏎- Team members

  6. #6 peak

    Teleostomi (teleost fish) > Osteichthyes (bony fish) > Class:⏎fish) > Stegocephalia (terrestrial⏎vertebrates) > Tetrapoda⏎(four-legged vertebrates) > Re

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:ThisIThisIItMaybeItMaybeItThisButITheItMaybeMaybeMaybeMaybeMaybeMaybeIThisButItButMaybeIThisTheThisButIThisIThisMaybeIThisThisMaybeThisThisIButIThisItI

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

    strengths and limitations of this theory are.⏎Ready to learn more about Mackinder’s Heartland Theory? Read on…⏎- What is Mackinder’s Heartland Theory?⏎- Why is the Heartland so important?⏎- What are the strengths of the

  2. #2 peak

    the four most common types of soil and what plants grow best in them, so that you can understand your garden better!⏎Check it out!⏎- What Is Soil?⏎- Mixtures And Compounds⏎- What Is Matter?⏎- Combinations And

  3. #3 peak

    row)⏎corresponds exactly to the order of data points in the actual data file.⏎

  4. #4 peak

    and each of the bytes checked on the the ptr array are interpreted as unsigned char for the comparison.⏎- Pointer to the block of memory where the search is performed.⏎- Value to be located. The value is passed

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. L14:56 strongly suppresses document ends and list line breaks
  2. L10:84 strongly suppresses
  3. L13:10 strongly excites
  4. L1:1279 suppresses
  5. L14:1067 suppresses
  6. L6:67 excites
  7. L5:331 suppresses
  8. L9:25 excites
  9. L9:605 excites
  10. L8:32 excites
  11. L3:98 excites
  12. L2:204 excites
  13. L10:259 excites
  14. L0:37 excites
  15. L14:51 excites colon that introduces a statement
  16. L6:216 excites
  17. L5:238 excites
  18. L14:348 suppresses
  19. L5:285 suppresses
  20. L8:88 excites line breaks inside wrapped sentences
  21. L9:2729 suppresses
  22. L4:105 excites
  23. L2:256 weakly suppresses
  24. L14:415 weakly suppresses

Feeds later layers

  1. L16:313 strongly suppresses
  2. L17:829 strongly suppresses
  3. L18:429 strongly suppresses
  4. L17:205 strongly suppresses
  5. L16:190 strongly suppresses end-of-document boundary token
  6. L16:359 strongly suppresses
  7. L18:5 strongly suppresses
  8. L16:1579 strongly suppresses
  9. L20:346 suppresses
  10. L19:187 suppresses
  11. L17:1073 suppresses
  12. L17:432 suppresses
  13. L21:130 suppresses
  14. L16:2348 suppresses line break after a section heading
  15. L17:921 suppresses
  16. L18:705 suppresses
  17. L16:1421 suppresses
  18. L18:572 excites
  19. L18:528 suppresses
  20. L18:886 suppresses
  21. L17:855 suppresses
  22. L16:442 suppresses
  23. L18:3042 suppresses line breaks in wrapped text
  24. L22:506 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.

4 values between -4.90 and -4.663 values between -4.66 and -4.422 values between -4.42 and -4.181 values between -4.18 and -3.953 values between -3.95 and -3.714 values between -3.71 and -3.477 values between -3.47 and -3.2310 values between -3.23 and -3.002 values between -3.00 and -2.766 values between -2.76 and -2.527 values between -2.52 and -2.293 values between -2.29 and -2.059 values between -2.05 and -1.814 values between -1.81 and -1.572 values between -1.57 and -1.344 values between -1.34 and -1.102 values between -1.10 and -0.863 values between -0.86 and -0.633 values between -0.63 and -0.393 values between -0.39 and -0.1578 values between -0.15 and 0.091609 values between 0.09 and 0.321400 values between 0.32 and 0.56635 values between 0.56 and 0.80102 values between 0.80 and 1.041 values between 1.27 and 1.511 values between 4.36 and 4.59offnegativepositiveoften

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. ⏎→→⏎
  3. :\/\/
  4. ⏎→→→→⏎
  5. ⏎→ ⏎
  6. ⏎ ⏎
  7. onyms
  8. _ ⏎
  9. ) ⏎
  10. __ ⏎

Suppresses

  1. 1
  2. (a
  3. beck
  4. (d
  5. B
  6. 传闻
  7. 以下
  8. 6
  9. (
  10. 4