Neuron Explorer

Qwen3-0.6B

zero digit inside a number

A neuron in layer 10 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 '0' written in the middle of a number, such as in page numbers, p-values, dates, and counts. Turning it up made the model produce unrelated, wandering text.

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

    could pave the way for all-optical circuits (M Mützel et al 2002 Phys. Rev. Lett. 88 083601-1).⏎The light field of an interference pattern can

  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

    of Muhlenberg College’s newest sustainability initiatives.⏎By: Grace Oddo '22 Wednesday, June 30, 2021 02:03 PM⏎Professor of Biology and Director of Sustainability Studies Rich N

  4. #4 peak

    9-0.965). Children were more vulnerable to rotavirus (p = 0.008, OR<|endoftext|>A Global Index of Sustainable Well-Being⏎Source: The

  5. #5 peak

    Central Business District as seen from the top of the Carlton Centre⏎|Municipality||City of Johannesburg|⏎|• Total||1.09 km2 (0.42 sq mi)|⏎|• Density||13,00

  6. #6 peak

    . Check out our full-length feature to find out more about these animals' astounding mental skills.<|endoftext|>- Cooking Time: 0⏎- Servings: 0⏎- Here are some recipes I have collected over the years and I am thrilled

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: which is the one thing you have. So, which is a thing. So, which is a thing. So, which is a thing. So, which is a thing. So, which is a thing. So, which is a

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

    of Biblical Interpretation found in the catalog.⏎April 1, 2003⏎by Baker Academic⏎Written in English⏎|The Physical Object|⏎|Number of Pages||180|⏎Hermeneutics gives the reader

  2. #2 peak

    short-listed, especially since they’re volatile. With short-selling, volatility gives investors a chance to make big money.⏎Short-selling is an investment practice where you borrow stocks and sell them at a loss in hopes the price will drop. The purpose

  3. #3 peak

    WHO⏎The number of deaths from measles has fallen by 79 percent worldwide since 2000, thanks mainly to mass

  4. #4 peak

    the 1990s – when rising North Sea gas production and regulatory changes led to a massive increase in gas fired generating capacity – meant power emissions fell by 45 million tonnes or 22% over the decade. It’s

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. L7:873 suppresses
  2. L9:977 suppresses the last digit of a decade year
  3. L9:657 excites
  4. L8:41 excites
  5. L8:20 excites
  6. L9:424 suppresses
  7. L3:651 excites zero in dates inside URLs
  8. L5:130 suppresses
  9. L8:318 suppresses
  10. L7:753 weakly excites
  11. L8:707 weakly suppresses
  12. L9:1531 weakly suppresses
  13. L4:326 weakly excites
  14. L9:1744 weakly suppresses
  15. L9:2319 weakly suppresses
  16. L0:208 weakly excites the digit zero inside codes or numbers
  17. L9:1032 weakly suppresses
  18. L4:1215 weakly excites
  19. L9:1348 weakly excites
  20. L8:660 weakly suppresses
  21. L7:2983 weakly excites
  22. L7:830 weakly excites
  23. L5:2102 weakly excites
  24. L9:135 weakly excites

Feeds later layers

  1. L18:726 suppresses
  2. L16:724 suppresses
  3. L11:553 suppresses
  4. L16:1333 suppresses 'two' after 'divided into' or 'there are'
  5. L12:713 suppresses
  6. L13:1479 suppresses
  7. L16:1420 suppresses
  8. L18:410 suppresses
  9. L16:1060 suppresses
  10. L14:1528 suppresses
  11. L19:723 weakly suppresses
  12. L11:2994 weakly excites
  13. L18:2831 weakly suppresses
  14. L22:332 weakly suppresses
  15. L18:1166 weakly suppresses
  16. L12:2434 weakly suppresses
  17. L16:2056 weakly suppresses
  18. L21:110 weakly suppresses
  19. L17:907 weakly suppresses
  20. L19:946 weakly excites
  21. L19:851 weakly suppresses comma inside numbers with thousands
  22. L11:502 weakly excites
  23. L20:1715 weakly suppresses
  24. L11:1427 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.

1 values between -1.87 and -1.772 values between -1.46 and -1.353 values between -1.35 and -1.252 values between -1.25 and -1.142 values between -1.14 and -1.044 values between -1.04 and -0.946 values between -0.94 and -0.8320 values between -0.83 and -0.7323 values between -0.73 and -0.6247 values between -0.62 and -0.5278 values between -0.52 and -0.42115 values between -0.42 and -0.31177 values between -0.31 and -0.21349 values between -0.21 and -0.101068 values between -0.10 and 0.001156 values between 0.00 and 0.11564 values between 0.11 and 0.21203 values between 0.21 and 0.3161 values between 0.31 and 0.4212 values between 0.42 and 0.528 values between 0.52 and 0.632 values between 0.63 and 0.731 values between 1.15 and 1.251 values between 1.46 and 1.561 values between 1.98 and 2.091 values between 2.09 and 2.191 values between 2.19 and 2.29offnegativepositiveoften

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. 興趣
  2. uja
  3. IEWS
  4. тра
  5. хот
  6. aly
  7. 投降
  8. 的兴趣
  9. obbies
  10. ��

Suppresses

  1. iteration
  2. looping
  3. rencontrer
  4. placeholder
  5. Iter
  6. -bit
  7. 舸
  8. cyan
  9. 环
  10. Diameter