Neuron Explorer

Qwen3-0.6B

spot right before a number

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 fires on the spot just before a number is about to appear, such as after 'the rate of' or 'from the age of' when the next word is a figure. Turning it up made the model write number-heavy, repetitive text, so it seems tied to predicting numbers.

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

    = 2 x 1/4 or 1/4 is 1/2 of 1/2<|endoftext|>Messier 52 (also known as M 5

  2. #2 peak

    3.6 years over last 36 years, from the age of 21.4 years in 1970 to the age of 25 in 2006.⏎Researchers T.J. Mathews and

  3. #3 peak

    9.3% in 1967 to 26.2% in 2014, which was still double the white rate of 12.7%⏎Education. In 1950 only 1

  4. #4 peak

    fact-family partner. If your student doesn’t recognize the logic of that switch, review how fact families work by using smaller numbers: 5 = 9 – 4 means also that 4 = 9 – 5.⏎Made =

  5. #5 peak

    women (10.0 per 100,000), with mortality rates twice as high for African Americans as Whites (4.4 versus 2.2 per 100,000).⏎These researchers

  6. #6 peak

    evergreen broadleaf forest. In this context, we observed the restoration dynamics of plant communities over a period of 15 years (1996 to 2011) with different levels of harvesting intensity, including selective harvesting at

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 be careful with the user's own and the user's own's's and the user's's's and the user's's's, because the user's's's, and the user's's, because the user's,

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

    caused<|endoftext|>This preview shows pages 1–2. Sign up to view the full content.⏎This preview has intentionally blurred sections

  2. #2 peak

    8 Counting and Cardinality Task Cards (CC) 9-13 Numbers and Operations in Base Ten Task Cards (NBT) 14-18 Operations and Algebraic Thinking Task Cards (OA) 19-2

  3. #3 peak

    no control. A quick check of Jesus’ parables shows the relationships between slaves/servants and wealthy people. (See Matthew 20:1-16, Luke 16:1-13 or Luke 19

  4. #4 peak

    available⏎Int. J. Environ. Res. Public Health 2012, 9(11), 4170-4186; doi:1

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. L13:2357 suppresses
  2. L13:2303 excites
  3. L13:1372 excites number before an age or week count
  4. L12:2516 suppresses
  5. L14:131 excites
  6. L13:2997 suppresses
  7. L14:2195 weakly excites
  8. L0:136 weakly suppresses space after a dash in number ranges
  9. L14:2078 weakly excites
  10. L8:7 weakly excites
  11. L12:698 weakly excites
  12. L11:2037 weakly suppresses
  13. L13:1308 weakly excites
  14. L6:756 weakly suppresses
  15. L8:16 weakly suppresses
  16. L14:1081 weakly suppresses
  17. L5:2642 weakly excites
  18. L14:926 weakly suppresses
  19. L9:2326 weakly suppresses
  20. L12:1269 weakly excites
  21. L13:573 weakly suppresses
  22. L6:1754 weakly excites
  23. L13:587 weakly excites
  24. L14:2494 weakly suppresses

Feeds later layers

  1. L16:1107 excites
  2. L18:987 excites
  3. L18:1904 excites
  4. L18:2084 excites
  5. L16:256 excites number right after 'age of' or 'ages'
  6. L17:2957 weakly excites
  7. L16:270 weakly excites
  8. L16:2048 weakly excites
  9. L17:712 weakly excites
  10. L16:2353 weakly suppresses
  11. L24:1460 weakly excites
  12. L19:1226 weakly excites
  13. L24:2931 weakly suppresses
  14. L20:1170 weakly excites
  15. L16:2105 weakly excites
  16. L20:1684 weakly excites
  17. L25:1830 weakly excites
  18. L17:577 weakly excites
  19. L17:676 weakly excites
  20. L17:2904 weakly suppresses
  21. L19:293 weakly suppresses
  22. L18:2152 weakly excites
  23. L18:629 weakly excites
  24. L19:2568 weakly excites

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.

2 values between -0.63 and -0.5810 values between -0.52 and -0.478 values between -0.47 and -0.4114 values between -0.41 and -0.3627 values between -0.36 and -0.3130 values between -0.31 and -0.2555 values between -0.25 and -0.20121 values between -0.20 and -0.14235 values between -0.14 and -0.09473 values between -0.09 and -0.041466 values between -0.04 and 0.02891 values between 0.02 and 0.07322 values between 0.07 and 0.13122 values between 0.13 and 0.1860 values between 0.18 and 0.2328 values between 0.23 and 0.2914 values between 0.29 and 0.349 values between 0.34 and 0.406 values between 0.40 and 0.451 values between 0.45 and 0.505 values between 0.50 and 0.563 values between 0.56 and 0.614 values between 0.83 and 0.881 values between 0.88 and 0.931 values between 1.47 and 1.53offnegativepositiveoften

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. uristic
  2. witnessing
  3. telesc
  4. Majesty
  5. �
  6. 辅助
  7. ведение
  8. thé
  9. agal
  10. aur

Suppresses

  1. ies
  2. amics
  3. Bush
  4. isan
  5. 骈
  6. OWN
  7. langs
  8. />⏎⏎
  9. cribes
  10. Exports