Neuron Explorer

Qwen3-0.6B

'of' linking a category to its parent field

A neuron in layer 6 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word 'of' (or 'within') when it links a specific thing to a bigger group it belongs to, like 'a subfield of algebra' or 'a subset of hash functions'. When turned up, the model's text drifted toward maths and abstract topics.

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

    classically Diophantine geometry), a subfield of algebraic number theory.⏎- The study of the real points of an algebraic

  2. #2 peak

    functions are generaelly a subset of hash functions that fulfil at least two extra criteria:⏎In order to fulfil these criteria

  3. #3 peak

    a subgroup of the N numbers.⏎Another mystery was solved in researching this class. At the Brid

  4. #4 peak

    .<|endoftext|>Archive for the ‘transcendental geometry’ tag⏎The modern mathematician finds the space of three dimensions, in which our visible universe is containled, entirely too contracted for his

  5. #5 peak

    based in the Northwestern United States, provided for U.S. government customers, complementing existing government agencies already using the US East Region. Each Region is wholly contained within a single country and all of its data and services stay within the designated Region.⏎

  6. #6 peak

    get to normal subduing the demand. The government starts to deepen the pocket by increasing taxes,

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: What is the name of the new type of car and what is the impact of the environmental movement? I don't know what to do with the old cars. The old cars have a very high impact on the environment. This is the case

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

    to explore STEM topics in a way that deepens their knowledge while also developing 21st-century skills.⏎Read the rest of this entry »⏎Filed under: Gaming, Services, Software, 3 million dollar grant, Associate Professor, Aug

  2. #2 peak

    in running a safer and quicker service.⏎- (Side effect) improved passenger movement will help to reduce larger crowds and subsequently improving safety.<|endoftext|>This article was submitted by a member of our State Climate Policy Network (SCPN) — a rapidly-growing

  3. #3 peak

    flanked by terraces for farming.⏎The stone for the buildings was mined from a quarry and shaped using bronze tools, then smoothed with sand in order to fit tightly together. No mortar was used in the construction. Even after many centuries of wear

  4. #4 peak

    bright, mocking beauty.⏎October 2003 | Volume 54, Issue 5⏎Your example of occupy brings to mind the way in which new words are

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. L4:2820 suppresses
  2. L5:60 excites
  3. L5:1477 excites
  4. L2:1036 excites
  5. L4:1738 weakly suppresses
  6. L2:1634 weakly suppresses
  7. L5:158 weakly suppresses
  8. L1:1464 weakly excites
  9. L1:1064 weakly suppresses
  10. L2:2692 weakly excites
  11. L5:1673 weakly suppresses
  12. L5:110 weakly suppresses
  13. L3:2282 weakly suppresses
  14. L3:2330 weakly suppresses
  15. L1:1225 weakly suppresses 'domain' after time or frequency
  16. L2:49 weakly suppresses
  17. L0:2748 weakly excites
  18. L1:2683 weakly excites
  19. L1:1852 weakly suppresses
  20. L2:596 weakly excites
  21. L5:86 weakly suppresses
  22. L2:2460 weakly excites
  23. L5:28 weakly excites
  24. L5:50 weakly excites

Feeds later layers

  1. L8:2256 excites
  2. L9:2566 excites
  3. L11:2292 weakly excites 'than' in comparisons
  4. L20:1481 weakly excites
  5. L19:2015 weakly excites
  6. L13:2573 weakly excites
  7. L17:1574 weakly excites
  8. L7:705 weakly suppresses
  9. L17:2959 weakly excites
  10. L10:2133 weakly excites
  11. L10:114 weakly suppresses
  12. L15:1355 weakly excites
  13. L10:544 weakly excites
  14. L23:2863 weakly suppresses
  15. L12:2367 weakly suppresses
  16. L15:2401 weakly suppresses
  17. L15:254 weakly excites
  18. L16:1618 weakly suppresses
  19. L8:2196 weakly excites comparison words like more, same, but
  20. L12:2289 weakly excites 'of' after part-whole nouns
  21. L9:2613 weakly suppresses
  22. L10:2966 weakly excites
  23. L7:1858 weakly excites
  24. L8:2070 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 -0.61 and -0.561 values between -0.56 and -0.524 values between -0.52 and -0.4810 values between -0.48 and -0.445 values between -0.44 and -0.4012 values between -0.40 and -0.3619 values between -0.36 and -0.3242 values between -0.32 and -0.2862 values between -0.28 and -0.2398 values between -0.23 and -0.19152 values between -0.19 and -0.15250 values between -0.15 and -0.11345 values between -0.11 and -0.07547 values between -0.07 and -0.03814 values between -0.03 and 0.01615 values between 0.01 and 0.06375 values between 0.06 and 0.10230 values between 0.10 and 0.14146 values between 0.14 and 0.1877 values between 0.18 and 0.2247 values between 0.22 and 0.2622 values between 0.26 and 0.3018 values between 0.30 and 0.345 values between 0.34 and 0.395 values between 0.39 and 0.432 values between 0.43 and 0.471 values between 0.47 and 0.512 values between 0.51 and 0.551 values between 1.00 and 1.05offnegativepositiveoften

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. isci
  2. โรงแรม
  3. 得住
  4. abinet
  5. criptive
  6. isi
  7. ligne
  8. nghi
  9. 躺着
  10. isiert

Suppresses

  1. puted
  2. mittel
  3. pdata
  4. ма
  5. ��이
  6. mitters
  7. newY
  8. etermination
  9. 税务总局
  10. letal