Neuron Explorer

Qwen3-0.6B

digits inside dates

A neuron in layer 20 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the number digits inside dates and timestamps, such as the zero in '04-06-2009'. Its negative side is less clear. Steering did not give a consistent picture.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    Ray, Viral Structure And Flu Vaccines.⏎04-07-2009 / The History Of Influenza⏎04-06-2009 / Flu Vaccine Contamination And Quality Control⏎0

  2. #2 peak

    information about the person's previous medical condition prior to death and so gives an understanding of the various factors which may have contributed to their death. Even if the cause of death may seem clear, the person may have had a medical condition which was not

  3. #3 peak

    9 / Preservative-Free Flu Shots⏎03-04-2009 / Flu Strains Developing Resistance to Tamiflu⏎03-03-2009 / The History Of Influenza⏎0

  4. #4 peak

    , everything should fit your needs.⏎Afterwards, do not forget to notice the condition and soil moisture. Whether the soil is fertile or not? If the soil is fertile, you don’t need to add the loose soil. However, if you

  5. #5 peak

    arthritis? It’s possible, according to researchers at Cardiff University in Wales who have been studying the Read more…⏎Posted in Jesus/Historical Jesus.⏎11/04 | BAR editor Hershel Shanks discusses a recent article in the Atlantic un

  6. #6 peak

    the cooking process.⏎Absolutely any food or dish that you would simmer in its liquid once you have brought it to the boil. All beans, legumes, grains and pulses; all soups

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: how many ways but not necessarily by by by by by by by by by by by by by by by by by by by by

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

    the OT? No, not according to Gordon Fee and Douglas Stuart in their popular book How to Read the Bible for All Its Worth. Following are ten principles for interpreting OT narratives that Fee and Stuart recommend:⏎

  2. #2 peak

    an individual and physical examination must be performed in a respectful and culturally sensitive manner. Privacy is essential and the nurse must be aware of their own ...⏎You have reached your viewing limit for this book.⏎What people are saying - Write a review⏎Res

  3. #3 peak

    www.gpi.org.⏎Answers: C, A, B, B, C<|endoftext|>This article is about the Great American Novel as a concept. For other uses, see Great American Novel (disambiguation).⏎The "Great American Novel"

  4. #4 peak

    !⏎What about NetBeans? A popular, widely used IDE that many large applications have been created in over the years. Do you know how to use it to its full potential? Well, in this course we will go through several lessons that will

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. L18:569 suppresses
  2. L17:43 suppresses
  3. L18:850 excites
  4. L19:2143 excites
  5. L19:423 suppresses words that set up a following clause
  6. L15:1662 suppresses
  7. L19:146 suppresses
  8. L18:2724 suppresses
  9. L2:1422 excites
  10. L16:918 excites
  11. L18:141 weakly excites
  12. L19:1653 weakly suppresses
  13. L7:805 weakly excites
  14. L19:1499 weakly excites
  15. L19:1119 weakly excites
  16. L19:1483 weakly suppresses
  17. L15:20 weakly excites
  18. L18:2149 weakly excites
  19. L18:1326 weakly excites
  20. L10:678 weakly suppresses
  21. L7:26 weakly suppresses
  22. L17:295 weakly suppresses
  23. L18:1927 weakly suppresses
  24. L19:988 weakly excites

Feeds later layers

  1. L24:363 suppresses
  2. L21:653 excites
  3. L26:511 suppresses
  4. L26:5 weakly excites
  5. L26:732 weakly suppresses
  6. L21:112 weakly excites
  7. L21:324 weakly excites
  8. L21:170 weakly excites
  9. L27:948 weakly excites
  10. L25:5 weakly excites
  11. L24:85 weakly excites
  12. L26:315 weakly suppresses
  13. L25:409 weakly suppresses
  14. L27:314 weakly suppresses
  15. L25:249 weakly suppresses
  16. L27:1236 weakly excites
  17. L27:127 weakly suppresses
  18. L27:425 weakly suppresses
  19. L27:1076 weakly excites
  20. L27:26 weakly suppresses
  21. L27:1 weakly excites
  22. L22:90 weakly excites
  23. L27:84 weakly suppresses
  24. L25:1114 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 -9.75 and -9.101 values between -8.44 and -7.793 values between -7.79 and -7.131 values between -7.13 and -6.481 values between -5.82 and -5.172 values between -5.17 and -4.519 values between -4.51 and -3.8611 values between -3.86 and -3.208 values between -3.20 and -2.5523 values between -2.55 and -1.8927 values between -1.89 and -1.2489 values between -1.24 and -0.58776 values between -0.58 and 0.07666 values between 0.07 and 0.73460 values between 0.73 and 1.38381 values between 1.38 and 2.03301 values between 2.03 and 2.69299 values between 2.69 and 3.34222 values between 3.34 and 4.00166 values between 4.00 and 4.65137 values between 4.65 and 5.3197 values between 5.31 and 5.9676 values between 5.96 and 6.6249 values between 6.62 and 7.2736 values between 7.27 and 7.9321 values between 7.93 and 8.5814 values between 8.58 and 9.249 values between 9.24 and 9.8910 values between 9.89 and 10.551 values between 10.55 and 11.202 values between 11.20 and 11.863 values between 11.86 and 12.513 values between 13.16 and 13.821 values between 14.47 and 15.132 values between 15.78 and 16.44offnegativepositiveoften

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. boa
  2. umb
  3. larger
  4. bigot
  5. 还将
  6. PIO
  7. resignation
  8. mode
  9. wag
  10. satisfactory

Suppresses

  1. ctors
  2. ainer
  3. ainers
  4. ointments
  5. ission
  6. verages
  7. itter
  8. efore
  9. missions
  10. imming