Neuron Explorer

Qwen3-0.6B

'of' after a noun phrase

A neuron in layer 17 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 "of" when it follows a noun that needs a following phrase, such as "one of the most unique caves of" or "a trading post of". Turning it down made the model repeat "of" over and over, so this neuron seems tied to the word "of".

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

    Here's a link to place your order.⏎The passage above needs a great deal of care and attention because this is one of the most difficult 'shifts' of 2nd finger to get in tune. When you are playing the tune you

  2. #2 peak

    AD and subsequently belonged to a series of local empires. Modern Singapore was founded in 1819 by Sir Thomas Stamford Raffles as a trading post of the East India Company with permission from the Johor Sultanate. The British obtained

  3. #3 peak

    . They have been reintroduced to National Reserves of Mongolia and China⏎CONSERVATION STATUS: Classified as Extinct in the

  4. #4 peak

    different strategies developed by elite men of color after the Haitian revolution to secure their status in the Atlantic World. Considered as “the first truly modern international crisis of exiles,” the

  5. #5 peak

    round Carolingian shields was the use of a sugarloaf boss in the centre which served to deflect the blows of an opponent and also as a pointed weapon of

  6. #6 peak

    the trauma of exile and one of the most devastating genocides of modern times.⏎Some 1.5 million Circassians were killed in the Caucasian War of the mid-to-late 19

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: of of of, or else? of or else? of or else? 你也可以吗? 里面我看看看看。 里面好像上你看看看看。 之间还是上一下。

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

    449F for pure metal and about 512F for tin-zinc alloys. Tests show that screws plated with the material do not stand much of a chance in hot areas. One test run several years ago by J. B

  2. #2 peak

    the language?⏎No, I’d say it started out as a Standard English euphemism. It wasn’t much of a semantic leap to take a word which literally means something that roosts and

  3. #3 peak

    sure that no Earthly gunk makes it off-world—primarily so that it doesn’t contaminate whatever science an Earth-born probe is doing.⏎None of that has ever really been a problem.

  4. #4 peak

    olysis of human red blood cells would occur if the cell were in .. solution A) isotonic B) hypertonic C) hypotonic D) none of these I think it's B. I'm not too sure though.⏎November

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. L16:1088 strongly excites 'of' after a noun phrase, as in 'of course'
  2. L3:315 excites 'of' after a noun phrase
  3. L15:785 suppresses 'of' in set phrases like 'of course'
  4. L4:4 suppresses
  5. L6:357 suppresses
  6. L5:363 excites
  7. L0:2726 excites
  8. L2:895 suppresses
  9. L0:1220 suppresses
  10. L0:40 suppresses 'of' in 'sort of' hedges
  11. L6:13 excites
  12. L9:2264 excites 'of' after superlatives and 'the X of'
  13. L10:180 excites
  14. L1:2567 excites
  15. L15:80 suppresses
  16. L10:469 suppresses 'of' after a noun phrase
  17. L16:428 weakly excites unclear: mid-sentence commas, 'most' before a verb
  18. L8:14 weakly excites
  19. L1:725 weakly suppresses "to" in "referred to"
  20. L16:41 weakly suppresses
  21. L5:1896 weakly suppresses
  22. L0:1224 weakly suppresses
  23. L13:132 weakly excites "of" after a noun
  24. L16:374 weakly excites

Feeds later layers

  1. L19:715 excites
  2. L21:2802 excites 'of' inside set phrases like 'course of'
  3. L18:2808 excites
  4. L20:64 suppresses
  5. L23:567 suppresses
  6. L20:2698 excites
  7. L22:160 excites
  8. L20:1972 excites 'of' in 'of course'
  9. L22:229 excites "of" after a noun in "X of Y"
  10. L25:286 excites
  11. L18:953 suppresses
  12. L19:1992 excites
  13. L20:2599 suppresses
  14. L21:2924 suppresses
  15. L25:2020 suppresses
  16. L23:128 excites 'for' as in 'for example'
  17. L18:2244 excites "of" inside titles and names
  18. L22:1993 suppresses
  19. L19:2507 suppresses
  20. L20:283 excites
  21. L22:2587 weakly excites
  22. L19:2088 weakly suppresses
  23. L22:288 weakly suppresses
  24. L18:2476 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.

73 values between -2.74 and -1.553072 values between -1.55 and -0.36607 values between -0.36 and 0.8427 values between 0.84 and 2.0310 values between 2.03 and 3.2213 values between 3.22 and 4.416 values between 4.41 and 5.605 values between 5.60 and 6.805 values between 6.80 and 7.993 values between 7.99 and 9.186 values between 9.18 and 10.376 values between 10.37 and 11.563 values between 11.56 and 12.765 values between 12.76 and 13.956 values between 15.14 and 16.336 values between 16.33 and 17.521 values between 17.52 and 18.724 values between 18.72 and 19.919 values between 19.91 and 21.107 values between 21.10 and 22.292 values between 22.29 and 23.485 values between 23.48 and 24.672 values between 24.67 and 25.874 values between 25.87 and 27.063 values between 27.06 and 28.251 values between 28.25 and 29.442 values between 29.44 and 30.633 values between 30.63 and 31.832 values between 31.83 and 33.021 values between 33.02 and 34.212 values between 34.21 and 35.403 values between 35.40 and 36.592 values between 36.59 and 37.792 values between 43.75 and 44.94offnegativepositiveoften

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. sorts
  2. appropri
  3. aths
  4. ự
  5. variables
  6. inclusion
  7. importance
  8. SCE
  9. circuits
  10. contr

Suppresses

  1. licate
  2. orial
  3. OfWork
  4. iotic
  5. →of
  6. eful
  7. liness
  8. headed
  9. aceous
  10. amework