Neuron Explorer

Qwen3-0.6B

'of' in 'of course'

A neuron in layer 20 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

Fires on the word 'of' inside the phrase 'of course', mostly when it is set off by commas in the middle of a sentence. The steering results are noisy and do not clearly show a theme.

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

    sent to our brain. Our brain then creates the beautiful pictures that we see before us. Our eyes are well-oiled machines (metaphorically speaking, of course).

  2. #2 peak

    do comment or send me an email @kcl.ac.uk. If I use it you will of course be credited by name in the essay although this will certainly not make you rich

  3. #3 peak

    of materials and weight and requires something that is much harder to come by: a propellant. You can make your own gunpowder, of course, but refining the process to create effective munitions is as tricky as building

  4. #4 peak

    opportunities or even just advice. They can connect⏎directly with employers and recruiters who can help them reach their goals faster than they ever thought possible. And, of course, Guru99 notes that online job boards function as the perfect resource for

  5. #5 peak

    introductory lesson I taught to our elementary children last Sunday ( 2/16/20). We had wonderful dialogue about the passages and stories below, and of course lots of other things! This followed a couple of weeks where we teachers taught

  6. #6 peak

    the belief that the intention of the author was the one and only correct interpretation of a work. Certainly, there can be other valid interpretations of a work, and of course the literature scholar should not be limited solely to the author, who may have

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: a person a a person a a a a a a a a a a a a a a a a a a a a the a a a a a a a a a the a a a a a a a the a a a a

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

    common are sinusitis, colds, allergies and a deviated septum. These conditions can cause frequent congestion, which can irritate the membranes and blood vessels of the nose. Medications that are used to treat conditions like sinusitis, cold

  2. #2 peak

    .⏎First, I want you to consider this story.⏎This story of American history has long been known as the “American dream” story, as the first immigrants of European descent came to this country

  3. #3 peak

    into eight sections. The pagoda is in the Burmese style. Burmese craftsmen were even brought in to carve all the "accessories" of the pagoda.<|endoftext|>Using electroshock therapy, scientists say it may be

  4. #4 peak

    collect and thus directly affect the quality of the image you see through the eyepiece.⏎The aperture is so paramount that some suppliers include it directly in the name of their telescope. For example, the PowerSeekr 70az from Cele

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. L19:1305 strongly suppresses unclear: verbs after a noun or 'that'
  2. L18:2271 strongly excites
  3. L15:632 excites
  4. L17:790 suppresses
  5. L17:264 suppresses
  6. L19:1881 suppresses
  7. L19:1314 excites
  8. L19:2449 suppresses
  9. L18:1601 suppresses
  10. L17:1167 excites
  11. L19:1988 suppresses
  12. L15:785 suppresses 'of' in set phrases like 'of course'
  13. L19:1024 suppresses
  14. L15:675 excites
  15. L19:1182 suppresses 'de' inside Spanish names
  16. L17:317 excites 'of' after a noun phrase
  17. L18:277 excites
  18. L19:766 excites
  19. L4:1686 excites
  20. L5:363 excites
  21. L15:1283 excites
  22. L18:520 excites
  23. L18:1588 excites
  24. L19:1434 suppresses

Feeds later layers

  1. L22:1124 strongly suppresses
  2. L22:360 suppresses
  3. L21:1457 suppresses
  4. L24:49 excites
  5. L25:247 suppresses
  6. L22:353 excites
  7. L23:24 excites
  8. L23:157 suppresses
  9. L27:108 excites
  10. L22:197 excites
  11. L26:672 suppresses
  12. L23:94 suppresses
  13. L25:525 suppresses
  14. L21:784 suppresses
  15. L21:1119 suppresses
  16. L22:324 suppresses
  17. L23:1180 suppresses
  18. L26:693 suppresses
  19. L21:920 suppresses
  20. L21:989 suppresses 'the' before a body part or organ
  21. L24:340 suppresses
  22. L27:121 excites
  23. L21:1050 suppresses
  24. L21:775 suppresses the article 'a' before a noun

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 -9.16 and -8.771 values between -8.77 and -8.381 values between -7.22 and -6.831 values between -6.83 and -6.442 values between -6.44 and -6.053 values between -6.05 and -5.661 values between -5.66 and -5.287 values between -5.28 and -4.894 values between -4.89 and -4.507 values between -4.50 and -4.1113 values between -4.11 and -3.7211 values between -3.72 and -3.3413 values between -3.34 and -2.9515 values between -2.95 and -2.5615 values between -2.56 and -2.1728 values between -2.17 and -1.7829 values between -1.78 and -1.4041 values between -1.40 and -1.0173 values between -1.01 and -0.62284 values between -0.62 and -0.231202 values between -0.23 and 0.161452 values between 0.16 and 0.55577 values between 0.55 and 0.93109 values between 0.93 and 1.329 values between 1.32 and 1.714 values between 1.71 and 2.101 values between 2.10 and 2.491 values between 2.49 and 2.872 values between 3.26 and 3.651 values between 5.98 and 6.37offnegativepositiveoften

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. 给她
  2. İçin
  3. 的文化
  4. 在这里
  5. 所以他
  6. nét
  7. 的压力
  8. Nimbus
  9. 对她
  10. 的艺术

Suppresses

  1. the
  2. 该县
  3. our
  4. each
  5. this
  6. 该院
  7. 各项
  8. 我国
  9. 该
  10. 这家