Neuron Explorer

Qwen3-0.6B

'by' introducing a method or means

A neuron in layer 17 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 word 'by' when it names how something is done, like 'sorted by length', 'by email', or 'by extension'. Turning it up gave wandering, repetitive sentences, which don't clearly show the pattern.

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

    we'll notify you by email when courses for each new term are displayed here.<|endoftext|>Epigenetics is

  2. #2 peak

    , history, and society.⏎About the authors⏎Elaine Treharne is Roberta Bowman Denning Professor of Humanities and Professor of English (and, by courtesy, of German Studies) at Stanford University, as well as Director of Stanford

  3. #3 peak

    focus" uses 5 letters: C F O S U.⏎No direct anagrams for focus found in this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with

  4. #4 peak

    ://nrich.maths.org/⏎is one of a series of problems designed to develop learners' team⏎working skills. Other tasks in the series can be found by going to⏎What are you aiming to⏎For the task:⏎The team has

  5. #5 peak

    version of bool-extras is 0.3.0 and is available on Hackage.⏎If you have cabal-install (which you really should have, by the way), the

  6. #6 peak

    eat is the main ingredient in your diet.⏎Food is food insofar as our tastes are concerned. But, when it comes to our stomachs, and, by extension, our whole body is involved, food is more than just food. It

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: The next time, from now on, the next time, the next time, and so on. The next time. So. in. on. from. . . . . from. . . . . from. . . . from

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

    making the subject seem baffling and inaccessible to the outsider. Thankfully, Patrick breaks down any barriers by describing the fundamentals extremely well. Important beginner questions are answered thoroughly from the start: why don’t people bother drawing carbon hydrogen bonds? What is a

  2. #2 peak

    all are waiting for. Teachers, parents and student, now you will find below the Solutions of Class 10 Science Chapter 4 Environmental management questions answers. From the subject English of Maharashtra State Board.⏎Below you can see

  3. #3 peak

    teachers that have interest in using the Haber video in their curriculum. Reading materials and lesson ideas are available. I recently used a lesson that my colleague created directly from the provided materials. You can check out that lesson here. The lesson included some

  4. #4 peak

    bathing to prevent rickets and leading interpreted discussions on pregnancy and birth control.<|endoftext|>Where’s The Harm – From Materials Processing?⏎Chemical processing: Chemical companies produce more than 1,0

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. L9:1340 suppresses
  2. L10:1054 suppresses 'from' after movement or transfer verbs
  3. L4:1394 excites
  4. L5:763 suppresses
  5. L4:2347 excites
  6. L8:2982 excites
  7. L6:1726 suppresses
  8. L16:1079 suppresses "by" after a past participle
  9. L16:2307 excites
  10. L10:371 weakly suppresses
  11. L6:2594 weakly excites 'with' after a verb of action
  12. L10:201 weakly suppresses the letter B in vitamin names
  13. L13:3044 weakly excites
  14. L4:1013 weakly suppresses
  15. L16:2120 weakly suppresses
  16. L0:2486 weakly excites
  17. L16:1847 weakly suppresses
  18. L7:1734 weakly excites
  19. L12:2882 weakly suppresses
  20. L11:591 weakly suppresses 'if' in conditional asides
  21. L4:1126 weakly excites
  22. L15:2443 weakly suppresses
  23. L2:1554 weakly excites
  24. L2:1848 weakly suppresses

Feeds later layers

  1. L19:2273 excites
  2. L18:2916 excites
  3. L23:2763 excites
  4. L19:2719 excites
  5. L18:1747 excites
  6. L18:2607 excites
  7. L18:2599 weakly suppresses
  8. L18:2244 weakly excites "of" inside titles and names
  9. L27:1645 weakly excites
  10. L24:1915 weakly excites 'in' after a noun phrase
  11. L21:225 weakly suppresses
  12. L21:871 weakly excites
  13. L19:1618 weakly excites
  14. L19:2869 weakly excites
  15. L20:2239 weakly excites
  16. L19:1874 weakly excites
  17. L27:1086 weakly suppresses
  18. L24:137 weakly suppresses
  19. L21:461 weakly excites
  20. L19:1564 weakly excites
  21. L18:1165 weakly excites
  22. L21:763 weakly suppresses
  23. L18:1346 weakly suppresses
  24. L18:1509 weakly excites

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.

133 values between -1.61 and -0.942098 values between -0.94 and -0.281605 values between -0.28 and 0.3923 values between 0.39 and 1.067 values between 1.06 and 1.7210 values between 1.72 and 2.391 values between 2.39 and 3.064 values between 3.06 and 3.723 values between 3.72 and 4.395 values between 4.39 and 5.052 values between 5.72 and 6.391 values between 6.39 and 7.052 values between 7.05 and 7.721 values between 7.72 and 8.382 values between 8.38 and 9.051 values between 9.05 and 9.721 values between 10.38 and 11.051 values between 11.71 and 12.382 values between 12.38 and 13.041 values between 15.71 and 16.371 values between 19.04 and 19.701 values between 19.70 and 20.371 values between 21.04 and 21.701 values between 23.70 and 24.371 values between 24.37 and 25.03offnegativepositiveoften

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. 这种方式
  2. virtue
  3. means
  4. gunfire
  5. sheer
  6. conventions
  7. pite
  8. intuition
  9. es
  10. convention

Suppresses

  1. cape
  2. ://${
  3. stakes
  4. kus
  5. stag
  6. ��
  7. scar
  8. ldr
  9. 那人
  10. aned