Neuron Explorer

Qwen3-0.6B

sentence end before a line break

A neuron in layer 7 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 full stop or question mark that is immediately followed by a line break, especially where one topic ends and an unrelated one begins, like the end of a question in a list of exercises. It goes negative on line breaks in the middle of a sentence. Turning it up made the model's writing shift into new topics, but the link is weak.

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

    and International Federation of Gynecology and Obstetrics staging systems, which now have a unified staging scheme for ovarian, tubal, and peritoneal carcinomas.⏎Peritoneal Cancer - PubMed⏎(1 days ago) Primary cancer

  2. #2 peak

    hips. There⏎are two rows of seats in the waiting room. Using the rules, where⏎are they all sitting? Can you find all the possible ways?⏎What is the best way to shunt these carriages so that each train⏎

  3. #3 peak

    a specified capacity or efficiency performance level. Life may be measured in cycles and/or years, depending on the type of service for which the cell or battery is intended.⏎Btu (British Thermal Unit) — The amount of heat required to raise the

  4. #4 peak

    indented with spaces to the right. The relationships between the different parts are carried out through indenting. No numbers, letters, or Roman numerals are needed.⏎How do you do an outline method?⏎How to take notes using the outline method

  5. #5 peak

    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

  6. #6 peak

    triangles are joined to form a⏎quadrilateral. What do you notice about this quadrilateral as the⏎dynamic image changes? Can you prove your conjecture?⏎It is impossible to trisect an angle using

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: here is a new thing for me, here is a new thing for me, here is a new thing for me, here is a new thing for me, here is a new thing for me, here is a new thing for me,

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

    that it helps with seed development, photosynthesis, and promotes the chemical change that⏎- Potassium’s primary purpose is to help with the overall growth of your⏎cannabisplants. Potassium is responsible for assisting plants in growing healthy roots

  2. #2 peak

    The ruling also secured equal pay for women doing exactly the same work as men in male roles—a landmark achievement.⏎‘Equal pay for work of equal value’ established⏎ACAC’s review of the 1969 ruling found that few women

  3. #3 peak

    orus or Potassium Problem?⏎The most common symptoms of phosphorus deficiency are the following:⏎Deficiency signs will most likely first appear on the lowest leaves of a⏎The first signals of this deficiency are the following:⏎- The branches will turn a

  4. #4 peak

    .” 13 And suddenly there was with the angel a multitude of the heavenly host praising God and saying,⏎14 “Glory to God in the highest,⏎and on earth peace among those with whom he is pleased!”⏎15 When

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. L6:216 excites
  2. L4:733 excites
  3. L4:183 excites
  4. L6:67 excites
  5. L4:1950 excites
  6. L1:1279 suppresses
  7. L5:1653 weakly suppresses
  8. L6:710 weakly excites
  9. L3:1093 weakly suppresses
  10. L3:356 weakly excites
  11. L4:194 weakly suppresses
  12. L1:2434 weakly suppresses
  13. L6:1026 weakly suppresses
  14. L6:626 weakly suppresses
  15. L6:1512 weakly suppresses
  16. L4:41 weakly suppresses
  17. L4:202 weakly excites
  18. L6:255 weakly suppresses
  19. L6:871 weakly suppresses
  20. L5:128 weakly excites
  21. L0:385 weakly suppresses
  22. L1:2408 weakly excites
  23. L5:3046 weakly excites
  24. L0:72 weakly suppresses

Feeds later layers

  1. L8:1973 suppresses
  2. L22:377 excites
  3. L8:538 excites
  4. L16:317 excites
  5. L20:268 excites
  6. L16:795 weakly excites
  7. L14:568 weakly excites
  8. L9:234 weakly suppresses line break inside hard-wrapped text
  9. L10:102 weakly excites
  10. L21:1164 weakly excites
  11. L13:1498 weakly excites end of a document boundary
  12. L27:765 weakly excites
  13. L15:366 weakly excites
  14. L10:1244 weakly excites
  15. L17:2350 weakly suppresses
  16. L10:1450 weakly suppresses
  17. L18:1782 weakly excites
  18. L10:627 weakly excites
  19. L20:460 weakly suppresses
  20. L24:303 weakly suppresses
  21. L16:442 weakly suppresses
  22. L22:476 weakly excites
  23. L24:598 weakly suppresses
  24. L8:362 weakly suppresses unclear: evaluative or explanatory phrase words

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 -1.91 and -1.821 values between -1.57 and -1.481 values between -1.31 and -1.221 values between -1.05 and -0.971 values between -0.88 and -0.802 values between -0.80 and -0.712 values between -0.71 and -0.623 values between -0.62 and -0.543 values between -0.54 and -0.459 values between -0.45 and -0.3717 values between -0.37 and -0.2891 values between -0.28 and -0.20299 values between -0.20 and -0.11804 values between -0.11 and -0.021398 values between -0.02 and 0.06736 values between 0.06 and 0.15337 values between 0.15 and 0.23137 values between 0.23 and 0.3231 values between 0.32 and 0.4115 values between 0.41 and 0.495 values between 0.49 and 0.583 values between 0.58 and 0.662 values between 0.66 and 0.754 values between 0.75 and 0.831 values between 1.18 and 1.261 values between 1.26 and 1.351 values between 1.35 and 1.432 values between 1.43 and 1.52offnegativepositiveoften

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. agas
  2. NOWLED
  3. esda
  4. 马拉
  5. 掏
  6. residency
  7. ADR
  8. anas
  9. acements
  10. FieldValue

Suppresses

  1. kit
  2. _net
  3. Cups
  4. »⏎⏎
  5. ớt
  6. square
  7. cctor
  8. ottle
  9. direct
  10. yık