Neuron Explorer

Qwen3-0.6B

digits inside numbers

A neuron in layer 12 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 a digit that sits inside a number, such as the 0 in '£0.00', the 0 in a storm code like '09L', or the 1 in '1000'. It is strongest on single digits in counts and codes. When turned up, the model's output did not show a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).

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

    2013 - Tropical Storm Humberto (09L) off West Africa⏎On September 11, 2013 Tropical Depression 09L had strengthened enough to become the first storm of the year to reach hurricane

  2. #2 peak

    [The Daily Journal]<|endoftext|>Should You Ask Your Children to Apologize?By Craig Smith | September 23, 2016 | 0 comments⏎When kids say sorry, are they learning a lesson or just parro

  3. #3 peak

    often referred to as the 16-bit era, and the previous generation as the 8-bit.⏎About Gaming Consoles⏎1 reply to this topic⏎0 user(s) are

  4. #4 peak

    digit of one of the numbers, and then sum the results together.⏎There is a single positive integer T on the first line of input (equal to about 1000). It stands for the number of expressions to follow. Each expression

  5. #5 peak

    to see the full course over view and dates of courses running in your area.⏎You have 0 courses in your shopping cart⏎Current total : £0.00 + VAT⏎Join our mailing list to receive our tech

  6. #6 peak

    reading this topic⏎0 members, 0 guests, 0 anonymous users<|endoftext|>The JCTC applauds the action by

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 can use a certain level of information in a certain way, but I can make sure that the user will have to follow certain conditions. In other words, I can make sure that the user will have to be informed of the rules,

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

    Children and Adolescents (2011). Expert panel on integrated guidelines for cardiovascular health and risk reduction in children and adolescents: Summary report. Pediatrics, 128(S

  2. #2 peak

    be well represented in the efforts currently under way to elucidate the molecular basis of the genetic susceptibility to alcoholism.(JAMA. 1992;268:1877-1882)<|endoftext|>The Imag

  3. #3 peak

    ar abscess, mastoiditis, and rheumatic fever in children: time trend analysis. BMJ. August 6, 2005;331:328–9.⏎Used with permission from

  4. #4 peak

    12). Social connections and the persuasiveness of viral campaigns in social network sites: Persuasive intent as the underlying mechanism. Journal of Marketing, 18(1), 39-53. http://dx.doi

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. L6:2938 suppresses
  2. L10:800 excites
  3. L11:1601 excites
  4. L11:1546 excites
  5. L9:2890 excites
  6. L10:810 suppresses
  7. L11:313 weakly excites
  8. L11:300 weakly suppresses
  9. L0:1301 weakly suppresses
  10. L2:1703 weakly suppresses
  11. L4:3023 weakly suppresses
  12. L11:1153 weakly suppresses
  13. L1:2536 weakly excites
  14. L9:726 weakly suppresses
  15. L10:1705 weakly excites
  16. L9:2211 weakly suppresses
  17. L3:464 weakly excites
  18. L10:509 weakly excites
  19. L11:18 weakly suppresses
  20. L11:553 weakly suppresses
  21. L10:926 weakly suppresses
  22. L11:1716 weakly excites
  23. L2:1698 weakly excites
  24. L8:2337 weakly suppresses

Feeds later layers

  1. L14:542 strongly suppresses
  2. L17:259 suppresses
  3. L19:395 suppresses
  4. L16:256 suppresses number right after 'age of' or 'ages'
  5. L15:646 suppresses
  6. L16:2735 suppresses
  7. L16:3057 suppresses
  8. L17:533 suppresses
  9. L14:759 suppresses
  10. L16:582 suppresses
  11. L18:170 suppresses
  12. L13:999 suppresses
  13. L25:182 suppresses
  14. L14:311 suppresses
  15. L15:1900 excites
  16. L16:286 suppresses
  17. L15:364 suppresses
  18. L19:672 suppresses
  19. L16:253 suppresses
  20. L16:210 excites
  21. L24:197 suppresses
  22. L21:175 suppresses
  23. L19:851 suppresses comma inside numbers with thousands
  24. L20:1567 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 -13.41 and -13.061 values between -5.55 and -5.203 values between -4.48 and -4.121 values between -3.77 and -3.411 values between -3.41 and -3.051 values between -3.05 and -2.691 values between -2.69 and -2.342 values between -2.34 and -1.983 values between -1.98 and -1.625 values between -1.62 and -1.274 values between -1.27 and -0.915 values between -0.91 and -0.5551 values between -0.55 and -0.192904 values between -0.19 and 0.16907 values between 0.16 and 0.5218 values between 0.52 and 0.88offnegativepositiveoften

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. powered
  2. lifted
  3. Code
  4. extremism
  5. intage
  6. lifting
  7. ismatic
  8. ثور
  9. shaking
  10. typeof

Suppresses

  1. ços
  2. olini
  3. apixel
  4. erten
  5. ert
  6. ertain
  7. olin
  8. 年由
  9. 日正式
  10. erset