Neuron Explorer

Qwen3-0.6B

'of' after a percentage or share

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 fires on the word 'of' when it follows an amount or share, such as '31 percent of', 'Fewer than 1% of' or '60 percent of'. It is less active on words like 'with' or 'to' that end a phrase. Turning it up gave replies with no clear link to the trigger.

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

    race were 0.89% of the population.⏎There were 46 households out of which 28.3% had children under the age of 18 living with them, 65.2% were married<|endoftext|>

  2. #2 peak

    the hard, cortical bone that makes up the outer portion of these bones.⏎Like OFD, adamantinoma is very rare. Fewer than 1% of all cancerous tumors that begin in the bone are adamantinomas.⏎These slow

  3. #3 peak

    33 percent) than were nonveterans (19 percent).⏎Among Gulf War-era II veterans in 2009, 31 percent of veterans aged 18 to 24 years were enrolled in school. As

  4. #4 peak

    nea include older age, male sex, and higher BMI. In a new study of 1448 people living with Parkinson's, 45% of participants experienced sleep apnea. In addition to the general risk factors, for people

  5. #5 peak

    erry, India. You may make a digital copy or printout of this text for your personal, non-commercial use under the condition that you copy this document without modifications and in its entirety, including this copyright notice.⏎

  6. #6 peak

    15 the number of mobile phone subscribers in North Korea reached 3.24 million, (about 13%) and that about 60 percent of the population in Pyongyang, the capital, between 20 and 50

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 been thinking about the problem of the way of using the software and the way of using the software and the way of using the software and the way of using the software and the way of using the software and the way of using the

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

    6⏎“Those things, which ye have both learned, and received, and heard, and seen in me, do: and the God of peace will be with you” — Philippians 4:9⏎The theory or norm of

  2. #2 peak

    voice of a giant.⏎When the people looked over the barriers to see the black, smoking pits where the lightning bolts had struck, they realized what could have happened to them if they had advanced up the slopes. All the people, awed at

  3. #3 peak

    (GNBDK)⏎Never do to anyone else anything that you would not want someone to do to you. “Do not drink so much wine that you get drunk, and do

  4. #4 peak

    arrival of Jesus. This is why it is so significant that the angels described Jesus’ birth as “peace on earth.”⏎9 And an angel of the Lord appeared to them, and the glory of the Lord shone around them, and they were

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. L12:1653 suppresses
  2. L9:1909 excites
  3. L12:850 excites
  4. L11:2207 suppresses "of" after a noun in trademark and title phrases
  5. L12:2680 suppresses
  6. L12:2215 excites 'of' after 'one' or 'some' (partitives)
  7. L9:2871 suppresses
  8. L11:2497 suppresses
  9. L12:2347 suppresses
  10. L12:2535 suppresses
  11. L10:272 suppresses
  12. L12:2875 suppresses
  13. L8:1003 suppresses "the" or "of" before a noun phrase
  14. L11:1174 suppresses
  15. L11:3015 suppresses
  16. L12:1040 excites
  17. L3:2989 excites
  18. L10:1121 weakly excites 'to' or verb after 'what' or 'whether'
  19. L6:1222 weakly excites
  20. L11:1861 weakly suppresses
  21. L6:487 weakly suppresses
  22. L12:1404 weakly excites
  23. L6:1735 weakly excites
  24. L3:2185 weakly suppresses

Feeds later layers

  1. L14:2861 excites 'of' after a percentage
  2. L14:672 suppresses
  3. L18:2598 weakly excites
  4. L18:224 weakly suppresses
  5. L14:1191 weakly excites
  6. L18:516 weakly suppresses
  7. L17:224 weakly suppresses
  8. L16:229 weakly suppresses
  9. L16:928 weakly excites
  10. L20:2161 weakly excites
  11. L17:2792 weakly excites
  12. L22:231 weakly excites
  13. L14:1317 weakly excites
  14. L17:1652 weakly excites
  15. L17:225 weakly suppresses
  16. L18:92 weakly suppresses
  17. L15:1369 weakly suppresses
  18. L20:900 weakly suppresses end-of-text right after cut-off text
  19. L20:1942 weakly excites
  20. L15:1771 weakly excites
  21. L17:2034 weakly suppresses
  22. L17:593 weakly excites
  23. L14:968 weakly suppresses
  24. L15:37 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 -2.54 and -2.421 values between -2.06 and -1.943 values between -1.58 and -1.464 values between -1.46 and -1.344 values between -1.34 and -1.229 values between -1.22 and -1.105 values between -1.10 and -0.9810 values between -0.98 and -0.868 values between -0.86 and -0.7423 values between -0.74 and -0.6227 values between -0.62 and -0.5042 values between -0.50 and -0.38107 values between -0.38 and -0.26247 values between -0.26 and -0.14980 values between -0.14 and -0.021618 values between -0.02 and 0.10405 values between 0.10 and 0.22189 values between 0.22 and 0.34113 values between 0.34 and 0.4643 values between 0.46 and 0.5830 values between 0.58 and 0.7014 values between 0.70 and 0.826 values between 0.82 and 0.948 values between 0.94 and 1.063 values between 1.06 and 1.182 values between 1.18 and 1.302 values between 1.42 and 1.541 values between 1.54 and 1.661 values between 1.66 and 1.781 values between 2.02 and 2.141 values between 2.14 and 2.26offnegativepositiveoften

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. ctions
  2. dge
  3. oren
  4. already
  5. d
  6. SEE
  7. BCE
  8. [list
  9. neau
  10. atab

Suppresses

  1. ously
  2. izzy
  3. ToSend
  4. uss
  5. _except
  6. identally
  7. ivant
  8. reuseIdentifier
  9. razil
  10. 颠