Neuron Explorer

Qwen3-0.6B

line break inside hard-wrapped text

A neuron in layer 9 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 falls in the middle of a sentence, which happens when text was copied from a page with fixed-width lines. The strongest negative cases are hyphenated pieces inside web addresses. Turning it up did not give a clear topic.

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 private foundations⏎had supported the development of a plethora of⏎curricula and approaches to instruction; these⏎led to such R&D organizations as the Biological⏎Sciences Curriculum Study, the Chemical Bond⏎Approach, and the Physical Science

  2. #2 peak

    Oder⏎River and assault Berlin. As this was happening, British and Canadian forces were slogging their way towards the Rhine (taking the majority of Western⏎Allied losses in

  3. #3 peak

    that under the British⏎Mercantilism and American imperialism in chapter 4. While the crops⏎could produce incredible yields when properly managed and with plenty of⏎water, doing so reduced the water table, added salt deposits to the⏎land

  4. #4 peak

    priestly hierarchy; no saints, martyrs, or even a⏎founder; no scripture or exegesis; no interest in education and art - in⏎fact, no positive, constructive aspect whatsoever. So vague and amorphous⏎

  5. #5 peak

    an event⏎known as a Lunar Eclipse, in which the Earth's circular⏎shadow can be seen to move across the face of the Moon. Because of the⏎orbital arrangements involved,

  6. #6 peak

    has said⏎that he would not have one Christian missionary in India if he rose to⏎power. And if the Zionist Jews became the governors of Palestine all missionary⏎work there

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: - - - - - - - - - - - - - - - - - - - - - - - -

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

    Jeremy Rosen on Rav Lichtenstein is available from http://jeremyrosen.blogspot.co.uk/2015/08/rav-lichtenstein-zl.html⏎Your comment will be posted after it is approved.⏎Leave a Reply

  2. #2 peak

    ://volcansvanuatueruptionsgb.blogspot.fr/2014/03/08032014-nevado-de-colima-cleveland.html and used this image from January 2nd. We

  3. #3 peak

    and other biological safety cabinet components protect.⏎Read More @ https://cmibloginsight.blogspot.com/2023/05/biological-safety-cabinets-ensuring.html<|endoftext|>World

  4. #4 peak

    the stream they have investigated. If you are interested in becoming a⏎TEAM volunteer and helping educate Marylanders about our natural⏎resources, visit our website at⏎www.dnr.maryland.gov/education/teamdnr and fill out

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. L7:74 strongly suppresses unclear: scattered mid-sentence tokens
  2. L8:380 strongly suppresses
  3. L4:156 strongly suppresses
  4. L4:32 strongly excites
  5. L7:1240 strongly excites
  6. L6:1374 suppresses
  7. L3:468 excites
  8. L8:538 excites
  9. L6:7 excites
  10. L5:1490 excites
  11. L8:7 suppresses
  12. L7:682 suppresses
  13. L5:1778 excites
  14. L8:3 suppresses
  15. L7:133 excites line break inside a sentence
  16. L4:138 suppresses
  17. L4:119 suppresses
  18. L3:225 suppresses
  19. L4:36 suppresses
  20. L8:4 suppresses
  21. L8:16 excites
  22. L2:126 suppresses
  23. L8:13 excites
  24. L7:122 suppresses

Feeds later layers

  1. L16:795 strongly suppresses
  2. L10:230 strongly excites
  3. L22:377 strongly suppresses
  4. L20:268 strongly suppresses
  5. L21:1164 strongly suppresses
  6. L10:102 strongly suppresses
  7. L14:568 strongly suppresses
  8. L16:317 strongly suppresses
  9. L14:2751 strongly suppresses
  10. L17:910 strongly suppresses
  11. L20:652 suppresses
  12. L17:1242 suppresses
  13. L19:1079 suppresses
  14. L20:320 suppresses
  15. L15:643 excites
  16. L12:1012 suppresses
  17. L15:985 suppresses line breaks inside wrapped text
  18. L10:2315 suppresses line break inside wrapped text
  19. L18:795 suppresses
  20. L15:1448 suppresses
  21. L25:135 excites
  22. L19:558 suppresses
  23. L19:2773 suppresses
  24. L15:337 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 -1.64 and -1.504 values between -1.50 and -1.354 values between -1.35 and -1.212 values between -1.21 and -1.0712 values between -1.07 and -0.9319 values between -0.93 and -0.7818 values between -0.78 and -0.6455 values between -0.64 and -0.5075 values between -0.50 and -0.35122 values between -0.35 and -0.21172 values between -0.21 and -0.07307 values between -0.07 and 0.08327 values between 0.08 and 0.22423 values between 0.22 and 0.36466 values between 0.36 and 0.50418 values between 0.50 and 0.65392 values between 0.65 and 0.79309 values between 0.79 and 0.93248 values between 0.93 and 1.08170 values between 1.08 and 1.22116 values between 1.22 and 1.3694 values between 1.36 and 1.5054 values between 1.50 and 1.6535 values between 1.65 and 1.7922 values between 1.79 and 1.9316 values between 1.93 and 2.089 values between 2.08 and 2.226 values between 2.22 and 2.363 values between 2.36 and 2.501 values between 2.50 and 2.651 values between 2.65 and 2.792 values between 2.93 and 3.071 values between 3.22 and 3.361 values between 3.65 and 3.791 values between 3.93 and 4.07offnegativepositiveoften

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. loquent
  2. moll
  3. {\
  4. entarios
  5. bilit
  6. {_
  7. izin
  8. `.
  9. adaptiveStyles
  10. 里面的

Suppresses

  1. ⏎
  2. ึ
  3. nels
  4. ⏎
  5. pund
  6. __(⏎
  7. éis
  8. ⏎
  9. AFE
  10. ',...⏎