Neuron Explorer

Qwen3-0.6B

'to' after 'matters' or 'means' phrases

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 the word 'to' in phrases about what matters or means something to someone, such as 'means so much to me' or 'doesn't matter to them'. Turning it up did not bring out this theme in the written text.

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

    this essay would be for her. She simply wrote the following from her heart:⏎Freedom is a word that has different meanings to everyone. So what does freedom mean to me? Well, it is the describing word for our country. It’s the

  2. #2 peak

    es⏎- Back hole, arse, butt, bussy⏎- Breasts, chest, breast tissue, boobs, tits⏎Sex can mean different things to different people. Sex can include vaginal, anal, oral, hand jobs or fing

  3. #3 peak

    why I have a choice every day. Freedom means so much to me. It means being able to have a choice.⏎Without freedom in my life

  4. #4 peak

    in the office. But if their standards aren’t met, they get upset. Perfectionists don’t care about a company’s goal. It doesn’t matter to them. So they’ll always put their own goals or motivations first.⏎Focusing

  5. #5 peak

    Susan L. Holdt, Bertrand Jacquemin and Céline Rebours<|endoftext|>Performing well in college means everything to college students.⏎So much so, that as many as 87% of

  6. #6 peak

    of the American economy. At the time, Nixon had brought support from the public and even from economists who thought it was the only way to deal with what looked to be out of control inflation. By 1974, though, 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 number of a month is a number, and the number of a month is a number, and the number of a month is a number, and the number of a month is a number, and the number of a month is a number

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

    David Dubin⏎- LENS Videos⏎- Contact Us⏎Archive for author David Dubin⏎Los Angeles based LENS Neurofeedback provider and blogger.⏎

  2. #2 peak

    on the trees and sang so sweetly that the children used to stop their games in order to listen to them. ‘How happy we are here!’ they cried to each other.⏎One day the Giant came back. He had been to visit his

  3. #3 peak

    in the mid-18thC. You can hear each of these influences in her Harpsichord Concerto in A Major, a piece likely written for her own performances sometime between 1760 and 1782

  4. #4 peak

    work various bits of scenery.⏎In the 18th century, opera's popularity spread to England. One of the first Italian-language operas to be written for the London stage was George Frideric Handel's Rinaldo. The

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. L6:2242 suppresses
  2. L4:1035 suppresses 'for' after verbs like 'fall' or 'click'
  3. L3:678 weakly suppresses
  4. L6:2496 weakly suppresses
  5. L5:2817 weakly excites
  6. L7:1090 weakly excites
  7. L8:16 weakly suppresses
  8. L5:967 weakly excites
  9. L6:673 weakly excites "to" after adjectives like obvious or meaning
  10. L6:2567 weakly excites "by" after a passive verb
  11. L8:2619 weakly excites
  12. L5:110 weakly suppresses
  13. L3:1126 weakly excites 'could' after a noun subject
  14. L6:2349 weakly excites
  15. L5:1495 weakly excites
  16. L5:349 weakly excites
  17. L8:2194 weakly suppresses
  18. L2:518 weakly suppresses
  19. L2:104 weakly excites
  20. L8:2501 weakly excites 'iv' inside herbivores, energy fuel words
  21. L5:1588 weakly excites
  22. L3:1513 weakly suppresses
  23. L7:2252 weakly suppresses
  24. L1:1007 weakly excites name pieces inside a multi-part word

Feeds later layers

  1. L17:2770 excites
  2. L13:1514 excites
  3. L11:361 excites
  4. L14:1508 excites
  5. L15:2606 excites
  6. L16:2208 excites
  7. L17:2197 excites
  8. L10:501 suppresses
  9. L13:2465 excites
  10. L12:2889 excites
  11. L12:1904 excites
  12. L17:2898 excites
  13. L12:2535 excites
  14. L12:1582 excites
  15. L14:2042 excites comma after 'For X,' introducing a viewpoint
  16. L14:1682 excites
  17. L14:2508 excites
  18. L17:2459 excites
  19. L14:2840 excites
  20. L17:3003 weakly excites
  21. L12:1632 weakly suppresses
  22. L18:2195 weakly excites
  23. L26:970 weakly excites
  24. L10:223 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 -2.14 and -1.951 values between -1.77 and -1.593 values between -0.86 and -0.687 values between -0.68 and -0.5099 values between -0.50 and -0.31589 values between -0.31 and -0.131886 values between -0.13 and 0.051054 values between 0.05 and 0.23218 values between 0.23 and 0.4233 values between 0.42 and 0.603 values between 0.60 and 0.782 values between 0.78 and 0.962 values between 0.96 and 1.152 values between 1.15 and 1.331 values between 1.33 and 1.512 values between 1.51 and 1.691 values between 1.69 and 1.871 values between 2.06 and 2.241 values between 4.43 and 4.611 values between 4.61 and 4.791 values between 4.97 and 5.16offnegativepositiveoften

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. etro
  2. czną
  3. vm
  4. },{"
  5. iet
  6. culo
  7. 玩家来说
  8. lsi
  9. VM
  10. commentators

Suppresses

  1. ividad
  2. nten
  3. 뿐
  4. aled
  5. nage
  6. 还有什么
  7. 극
  8. 没问题
  9. 되었다
  10. 骄