Neuron Explorer

Qwen3-0.6B

unclear: 'or something' and open parentheses

A neuron in layer 19 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

The strongest triggers are phrases like 'or something' and an opening bracket after a name or term. The negative side is verbs like 'called' or 'referred' that come before a name. The pattern is mixed, and steering did not produce a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    16 25.........⏎Basics: Divide integers to get integer outputs in all cases⏎Determine value using Look-up table method⏎Create Volcano (or Atoll) martix. It is an extension of Bullseye matrix

  2. #2 peak

    't see any code to exit out of the loop, IE it will continue to call it self over and over until the program finally kills the heap/stack or something. I don't quite know what it is...⏎Also, even if it wasn

  3. #3 peak

    from each protein. They all have to agree in their "chemical properties" - if you try to pair up a hydrophobic with a charged amino acid or something like that, the association will be greatly weakened or eliminated.⏎Some biological structures (

  4. #4 peak

    (motivation)⏎• High levels of creativity⏎EXP: Many gifted students find logical reasoning to be their preferred learning style, & for others linguistics may be (Language Arts & Math are not automatically synonymous – as if they’re gifted in one

  5. #5 peak

    the⏎major arc AB,⏎where you go the long way around the bottom of the circle. Unless stated otherwise, it always means the⏎minor arc - the shortest of the two.⏎If you want to indicate the major arc, add

  6. #6 peak

    evidence: verbal comments or remarks. For example, calling a woman "aggressive," but a man "a go-getter" could be interpreted as discrimi...(register to read more)<|endoftext|>How Does Fake News Work?⏎What is Fake News

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 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

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

    monetary values of natural resources using an accounting system, or often referred to as Natural Capital Accounting, or NCA. Through NCA, natural resources

  2. #2 peak

    as its related contextual conditions. Although no single definition<|endoftext|>Early Church Views on Paul⏎1. Ebionites -- The First Christians⏎The earliest Christians were commonly called Ebionites, meaning "the Poor."⏎In G. Uhlhorn,

  3. #3 peak

    light that would pass through human tissue and to reveal details of the bones and tissue underneath. Because Roentgen didn’t know what the rays were, he called them “x”, meaning “unknown”.⏎The medical community and the public alike were

  4. #4 peak

    agreed to a Seagrass Charter drafted by Professor Mike Fortes to encourage countries to consider the values of seagrass when managing coastal environments.⏎This has been called the Bolinao Declaration and reads:⏎We, participants of the third International Se

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. L18:2981 suppresses
  2. L17:2844 suppresses
  3. L12:2252 suppresses
  4. L4:2639 suppresses
  5. L16:2289 excites 'or' linking two alternatives
  6. L8:1219 suppresses
  7. L17:2445 suppresses
  8. L4:860 weakly excites
  9. L14:1172 weakly excites
  10. L4:368 weakly suppresses
  11. L6:1930 weakly suppresses
  12. L5:330 weakly excites opening parenthesis after a noun
  13. L18:1735 weakly suppresses
  14. L18:1844 weakly suppresses
  15. L15:1223 weakly excites
  16. L16:2661 weakly suppresses
  17. L13:1336 weakly excites
  18. L14:736 weakly excites
  19. L18:638 weakly suppresses
  20. L13:1181 weakly excites 'of' after a definition or noun
  21. L5:2125 weakly suppresses
  22. L12:2203 weakly excites
  23. L18:2675 weakly suppresses
  24. L14:2910 weakly excites

Feeds later layers

  1. L21:945 excites
  2. L22:684 suppresses
  3. L20:955 suppresses
  4. L21:2900 suppresses
  5. L20:382 weakly suppresses
  6. L20:570 weakly suppresses
  7. L21:293 weakly suppresses
  8. L22:2578 weakly excites
  9. L21:863 weakly suppresses
  10. L22:2721 weakly excites
  11. L25:2661 weakly suppresses
  12. L21:419 weakly suppresses
  13. L26:588 weakly suppresses
  14. L21:7 weakly suppresses mixed, weak code and reference hits
  15. L20:1372 weakly suppresses
  16. L21:404 weakly suppresses
  17. L22:345 weakly suppresses
  18. L21:2474 weakly excites
  19. L25:2463 weakly suppresses
  20. L21:1265 weakly suppresses
  21. L20:254 weakly suppresses
  22. L23:24 weakly excites
  23. L21:793 weakly excites
  24. L26:2890 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.

1 values between -18.64 and -18.091 values between -17.00 and -16.451 values between -15.90 and -15.351 values between -15.35 and -14.811 values between -14.81 and -14.261 values between -13.71 and -13.161 values between -12.07 and -11.521 values between -10.97 and -10.421 values between -9.33 and -8.781 values between -8.78 and -8.231 values between -8.23 and -7.681 values between -7.68 and -7.144 values between -6.59 and -6.041 values between -6.04 and -5.495 values between -5.49 and -4.941 values between -4.94 and -4.402 values between -4.40 and -3.854 values between -3.85 and -3.303 values between -3.30 and -2.7510 values between -2.75 and -2.2010 values between -2.20 and -1.6613 values between -1.66 and -1.1144 values between -1.11 and -0.56784 values between -0.56 and -0.012624 values between -0.01 and 0.54373 values between 0.54 and 1.0816 values between 1.08 and 1.631 values between 1.63 and 2.181 values between 2.73 and 3.28offnegativepositiveoften

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. AtPath
  2. iator
  3. idge
  4. .getColumnIndex
  5. 的例子
  6. 但是对于
  7. hotmail
  8. 名叫
  9. sensed
  10. holster

Suppresses

  1. 作
  2. '])) ⏎
  3. *b
  4. as
  5. デ
  6. 作了
  7. Tr
  8. 做
  9. \'
  10. 'B