Neuron Explorer

Qwen3-0.6B

unclear: repeated or split word pieces

A neuron in layer 16 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron is hard to read. Its strongest moments land on a word fragment or a repeated word, like the "qu" in "askutasquash" or the second "practice". Its most negative moments are on words like "that" and "summer". Turning it up or down did not show a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).

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

    10,000 years to Mexico and America. The name squash, which means uncooked or eaten raw, comes from the Native American word askutasquash.⏎Butternut squash tastes fairly mild, somewhat sweet, and somewhat nut

  2. #2 peak

    so that each can hear what the other is really trying to express and not just what they expect or want to hear.⏎Listening is a skill that takes practice, practice, practice! The good news is that we can get better at it.⏎What

  3. #3 peak

    outlines why labels are so important, and explains how to read them. It also provides some examples of appropriate medicated feed labels.<|endoftext|>Sentences with Failure, Failure in a Sentence in English, Sentences For Failure⏎1. The business was

  4. #4 peak

    1037th UGM doctoral graduate.⏎In conclusion of her dissertation entitled "Biosystematics of Murricata L. Annona, Annona squamosa and Annona Reticulata by Numerical Approach

  5. #5 peak

    more of it! But in a few days, dopamine spikes no longer came in response to the grape: The brain doesn’t waste dopamine on the same-old, same-old, but saves it for new information about rewards.⏎

  6. #6 peak

    Causal clauses are those that tell the reason why the main clause happened or was done. In Latin, these are introduced by quod, quia, and quoniam (among others). This video discusses when the indicative mood is used in

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: I don't know of any. I don't know. I don't think I can't. I don't know. Maybe I should not be sure. I don't think of it as a question. I don't know. 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

    later hold a pencil for writing!⏎Why are art activities important?⏎Dr. Montessori once said, “Our schools prepare an eye that sees, a hand that obeys, and a soul that feels.” Through art activities, children learn how

  2. #2 peak

    Holes and gaps in the wall, roof, foundation, doors and windows allow air loss, which results in winter heat loss and summer heat gain. Tight construction, good insulation and high-performance windows are key. "

  3. #3 peak

    two classes, default on loan and no default on loan. There can also be multi-class classification which for example, is classifying

  4. #4 peak

    :⏎Time with you. By keeping your bedtime routine alive, you and your child get to do something new together — cheer for the good guys and boo the bad ones in the books you read. You also get a peek into how your

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. L15:1443 excites
  2. L14:1631 excites
  3. L15:3058 suppresses
  4. L15:9 excites
  5. L8:512 weakly suppresses
  6. L14:982 weakly excites
  7. L14:2152 weakly suppresses
  8. L15:7 weakly excites
  9. L14:1010 weakly excites
  10. L15:127 weakly excites
  11. L14:2851 weakly suppresses
  12. L6:1203 weakly excites
  13. L14:748 weakly excites
  14. L13:2053 weakly suppresses
  15. L13:813 weakly excites
  16. L10:1126 weakly suppresses
  17. L10:641 weakly excites
  18. L10:163 weakly suppresses
  19. L14:1033 weakly excites
  20. L14:669 weakly suppresses
  21. L3:541 weakly excites the word 'a' after play or places
  22. L13:2069 weakly excites
  23. L7:2155 weakly suppresses
  24. L15:1128 weakly excites

Feeds later layers

  1. L17:327 suppresses
  2. L17:1002 excites
  3. L17:220 suppresses
  4. L20:2552 excites
  5. L17:1488 excites
  6. L19:211 suppresses
  7. L18:2520 suppresses
  8. L19:1037 suppresses
  9. L20:1910 excites
  10. L17:1567 excites
  11. L18:959 excites
  12. L17:459 suppresses
  13. L19:1019 weakly excites
  14. L18:1083 weakly excites
  15. L17:1096 weakly suppresses
  16. L18:468 weakly suppresses
  17. L19:1110 weakly suppresses
  18. L19:911 weakly excites
  19. L19:711 weakly excites
  20. L21:300 weakly suppresses
  21. L19:1563 weakly suppresses
  22. L19:431 weakly excites
  23. L19:2657 weakly excites
  24. L17:920 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 -5.54 and -5.351 values between -5.16 and -4.971 values between -4.78 and -4.591 values between -3.63 and -3.441 values between -3.25 and -3.062 values between -3.06 and -2.873 values between -2.87 and -2.681 values between -2.68 and -2.483 values between -2.29 and -2.104 values between -2.10 and -1.913 values between -1.91 and -1.723 values between -1.72 and -1.534 values between -1.53 and -1.348 values between -1.34 and -1.1516 values between -1.15 and -0.958 values between -0.95 and -0.7616 values between -0.76 and -0.5730 values between -0.57 and -0.38146 values between -0.38 and -0.191572 values between -0.19 and 0.001481 values between 0.00 and 0.19308 values between 0.19 and 0.39144 values between 0.39 and 0.5875 values between 0.58 and 0.7734 values between 0.77 and 0.9617 values between 0.96 and 1.1512 values between 1.15 and 1.342 values between 1.34 and 1.533 values between 1.53 and 1.725 values between 1.72 and 1.913 values between 1.91 and 2.11offnegativepositiveoften

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. 因为在
  3. ~~~~~~~~~~~~~~~~
  4. abruptly
  5. 派驻
  6. 要把
  7. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  8. teenth
  9. 将自己的
  10. 一直都

Suppresses

  1. oph
  2. avs
  3. 克
  4. den
  5. 易
  6. sno
  7. lo
  8. 佛
  9. fish
  10. Dog