Neuron Explorer

Qwen3-0.6B

'to' before a verb after a noun

A neuron in layer 13 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 lights up on the word 'to' when it introduces a verb after a noun or adjective, as in 'no advantage to having', 'a way to train' or 'fun to research'. Turning it up gave text that drifted into unrelated general sentences, so the steering did not show the same 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

    You may decide you have to read the entire passage to answer the context questions.)Staying calm is always easier said than done but remember that there is no advantage to having existing subject knowledge on the passage. It’s kind of like an editor,

  2. #2 peak

    so on are a convenient way to prioritize everyday things. You get responsibilities in school done and you have a handy gadget.⏎What Others Are Reading Right Now.⏎How to Turn (Almost) Every Lady’s Head⏎Top female stylists share their favorite

  3. #3 peak

    to stay alive. At first no one had much choice in the matter; too few clergy were willing to make themselves known again. But as it became ever safer to be Christian, the problem came to a head. Adding to it was the problem

  4. #4 peak

    spiritual, and physical exercises that originated in India. It was created to control the mind and recognize the unaltered Witness Consciousness’. It is an ancient way to train oneself by practicing meditation, self control, yoga postures

  5. #5 peak

    7 graves. The result should be about 40. 000 graves.⏎You do not have the required permissions to view the files attached to this post.<|endoftext|>This overview outlines the tasks performed by

  6. #6 peak

    can communicate a thought, but your thought may not be understood. In some cases, your thought may not even reach the proper target. Thats why it pays to ask questions to make certain that people understand what you are saying. The great movie

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 have to a lot of people's support, especially in the community, especially if some people's support for the community's needs, especially in the community, especially in the situation in which they have been the most prominent role in the most

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

    mynursingpaper.com/wp-content/uploads/2020/06/cropped-1559720319884

  2. #2 peak

    ized or lack thereof).Please provide resources as required by APA standards.⏎https://mynursingpaper.com/wp-content/uploads/2020/06/cropped-1559720319884

  3. #3 peak

    ales as marine ecosystem engineers, Front Ecol Environ, http://www.joeroman.com/wordpress/wp-content/uploads/2014/09/roman-frontiers.pdf.<|endoftext|>Oil & Natural Gas Projects⏎Transmission, Distribution

  4. #4 peak

    election, incumbent Clark had a 30 percentage-point lead in campaign polls, but he lost to Republican Roger Jepsen, 52

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. L6:1133 strongly excites "to" before an infinitive verb
  2. L6:2747 suppresses
  3. L6:2064 suppresses
  4. L8:94 excites
  5. L10:616 excites
  6. L9:226 suppresses 'to' starting an infinitive
  7. L7:8 excites
  8. L9:1890 excites
  9. L6:1880 excites "by" introducing a deadline or means
  10. L8:369 suppresses 'of' after 'learned' or 'because'
  11. L9:330 suppresses gap where a number was removed
  12. L11:2064 excites
  13. L8:723 suppresses
  14. L9:1258 suppresses
  15. L12:1168 excites
  16. L8:1965 excites
  17. L3:1152 suppresses
  18. L9:288 suppresses
  19. L7:681 excites 'from' after prevent or stop
  20. L8:1546 excites
  21. L6:2359 excites
  22. L6:2000 suppresses
  23. L6:914 excites
  24. L6:2287 weakly suppresses

Feeds later layers

  1. L15:1176 excites
  2. L17:1592 excites
  3. L16:1153 excites
  4. L18:472 excites
  5. L15:1234 excites
  6. L19:385 suppresses
  7. L19:776 excites
  8. L17:472 excites
  9. L21:239 suppresses
  10. L24:208 excites
  11. L17:187 excites
  12. L20:241 excites
  13. L19:652 excites
  14. L15:448 suppresses
  15. L20:1214 excites
  16. L18:715 excites
  17. L24:265 excites
  18. L19:617 excites
  19. L27:401 excites
  20. L14:313 suppresses
  21. L19:1608 excites
  22. L25:181 excites
  23. L19:1126 excites
  24. L16:559 excites weak, mixed: verbs and adverbs in claims

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 -1.99 and -1.771 values between -1.77 and -1.552 values between -1.33 and -1.117 values between -1.11 and -0.8915 values between -0.89 and -0.6732 values between -0.67 and -0.45281 values between -0.45 and -0.221542 values between -0.22 and -0.001688 values between -0.00 and 0.22194 values between 0.22 and 0.4453 values between 0.44 and 0.6615 values between 0.66 and 0.8817 values between 0.88 and 1.1010 values between 1.10 and 1.3311 values between 1.33 and 1.554 values between 1.55 and 1.775 values between 1.77 and 1.995 values between 1.99 and 2.216 values between 2.21 and 2.434 values between 2.43 and 2.653 values between 2.65 and 2.872 values between 2.87 and 3.103 values between 3.10 and 3.324 values between 3.98 and 4.201 values between 4.42 and 4.641 values between 4.64 and 4.871 values between 6.63 and 6.86offnegativepositiveoften

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. odies
  2. deliver
  3. replace
  4. ãng
  5. uem
  6. create
  7. ODY
  8. 曉
  9. ilater
  10. ều

Suppresses

  1. 分歧
  2. invaluable
  3. 劣势
  4. 菜肴
  5. 尤为
  6. rollable
  7. 站长
  8. :+:
  9. 队长
  10. vocalist