Neuron Explorer

Qwen3-0.6B

'of' after a person noun

A neuron in layer 15 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 'of' when it follows a word for a person, like 'a friend of', 'a comrade of' or 'cousin of'. Turning it up made the model repeat 'of mine', 'of yours' and 'of hers', so it seems to help the model use possessive 'of' phrases.

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

    . Whenever he noticed a peasant beating his horse, Bergh would order the peasant to stop beating it at once.⏎On his way home from Russia, this friend of animals stopped in England, where he observed the work of the Royal Society for the

  2. #2 peak

    river and done trading until that Regiment was relieved by a South Carolina Regiment. The next morning as some of our men went to wash as usual, a comrade of Co. G was shot in the leg which was a signal for hostilities when

  3. #3 peak

    arring exists on this skin of the leg. Generally, weaker chemicals are used for the legs as they tend to

  4. #4 peak

    friend of Jesus who raised him from the dead (John, Chapters 11 &

  5. #5 peak

    friend of Allah, and one who does not work is considered by Allah, to be His

  6. #6 peak

    , France was plunged into a succession crisis over which distant cousin of the king would inherit the throne. Militarily and diplomatically, Francis'

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: It's not a bad idea to get a new idea in a new way to help help help...?... It... It... it... it... it... it... it... it... it... it... it... it... it...

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

    This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: itself, and if ( ) ( ) lim lim x c x c h x

  2. #2 peak

    2. Sign up to view the full content.⏎This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: 60 | | | | CHAPTER 1 FUNCTIONS AND MODELS Not all

  3. #3 peak

    concrete permeability⏎* Increases concrete strength⏎* Improves resistance to corrosion<|endoftext|>Prometheus shines brightly in this image, taken as part of the ongoing campaign to

  4. #4 peak

    dots” in a real-life political, often international scenario to create a plan of action.⏎You may be familiar with the word מוֹדִיעִין, also referring to intelligence (as well as to a city situated

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. L14:1807 excites
  2. L10:39 weakly excites
  3. L4:619 weakly excites
  4. L8:1100 weakly excites
  5. L11:199 weakly excites
  6. L5:2217 weakly suppresses
  7. L3:1110 weakly suppresses
  8. L4:2319 weakly excites
  9. L9:17 weakly excites
  10. L0:707 weakly excites
  11. L2:370 weakly excites
  12. L7:130 weakly suppresses
  13. L13:122 weakly excites
  14. L9:2202 weakly suppresses
  15. L14:943 weakly suppresses
  16. L4:705 weakly suppresses
  17. L8:10 weakly excites
  18. L2:2022 weakly excites
  19. L14:2034 weakly excites
  20. L0:2533 weakly excites
  21. L0:1754 weakly excites
  22. L4:2068 weakly suppresses
  23. L8:14 weakly excites
  24. L3:1912 weakly suppresses

Feeds later layers

  1. L17:2185 excites
  2. L16:961 excites
  3. L17:2736 excites
  4. L17:1771 excites
  5. L26:711 excites 'years' after an age number
  6. L19:2304 excites
  7. L27:447 excites
  8. L20:450 excites
  9. L25:2020 excites
  10. L22:142 suppresses 'of' after a quantity or noun phrase
  11. L23:567 excites
  12. L27:658 weakly excites
  13. L18:953 weakly excites
  14. L23:258 weakly excites
  15. L22:1993 weakly excites
  16. L17:488 weakly excites
  17. L18:2510 weakly excites
  18. L18:2693 weakly excites
  19. L24:2771 weakly excites
  20. L18:636 weakly suppresses
  21. L24:1650 weakly suppresses
  22. L19:3054 weakly excites
  23. L22:2457 weakly excites
  24. L22:2697 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 -1.20 and -1.106 values between -1.10 and -1.008 values between -1.00 and -0.907 values between -0.90 and -0.7914 values between -0.79 and -0.6923 values between -0.69 and -0.5952 values between -0.59 and -0.4992 values between -0.49 and -0.38164 values between -0.38 and -0.28337 values between -0.28 and -0.18708 values between -0.18 and -0.071599 values between -0.07 and 0.03654 values between 0.03 and 0.13192 values between 0.13 and 0.2436 values between 0.24 and 0.348 values between 0.34 and 0.441 values between 0.44 and 0.543 values between 0.54 and 0.651 values between 0.65 and 0.751 values between 1.06 and 1.161 values between 2.80 and 2.91offnegativepositiveoften

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. mine
  2. hers
  3. ours
  4. of
  5. theirs
  6. yours
  7. mine
  8. của
  9. Mine
  10. _of

Suppresses

  1. FileType
  2. 不限
  3. 作为一个
  4. 是一个
  5. глав
  6. Wrapped
  7. фи
  8. 落到
  9. ancode
  10. entials