Neuron Explorer

Qwen3-0.6B

the digit 3 in day-of-month dates

A neuron in layer 25 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 '3' in dates like 'June 30' or 'Feb. 3', which is the first digit of a day of the month. It is turned off by digits that start ages like '17' or '18'. Turning it up made the model talk about 'points' and repeat itself, and turning it down made it repeat the digit 8.

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

    atorsA student-built pollinator garden is one of Muhlenberg College’s newest sustainability initiatives.⏎By: Grace Oddo '22 Wednesday, June 30, 2021 02:03 PM⏎Professor

  2. #2 peak

    report go to the NOAA news website www.noaanews.noaa.gov⏎– Erich Luening⏎Print this page<|endoftext|>Today is Sunday, Feb. 3, the 34th day of 2013. There are

  3. #3 peak

    part of a series of features highlighting freshman seminars offered at Princeton this semester. Other features:⏎Exploring the science and nuance of facial perception⏎Posted October 31, 2011; 12:00 p.m

  4. #4 peak

    stakeholder groups.⏎Disseminate results to stakeholders and seek funding to develop a technology-based prototype⏎July 1, 2008 - June 30, 200

  5. #5 peak

    lower-income or urban schools on average already have a higher body mass index (BMI) and consume more soda. Another new study by Massachusetts General Hospital released June 3, indicates the pattern of fast food and obesity starts before they even begin school.⏎

  6. #6 peak

    - proper n. The northern parts of Scandinavia and Finland with the Kola Peninsula in Russia, traditionally inhabited by the Sami people.⏎from WordNet 3.0 Copyright 2006 by Princeton University

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:777688777777777777777777777777777777777777777777

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 Illinois Senate.Biography Roberts was born August 20, 1866 in Decatur, Michigan. He graduated from high school at 17 and became a teacher. He then chose to take Ph.D. coursework at

  2. #2 peak

    applauds the action by the FDA to monitor e-cigarettes, cigars, hookah tobacco and pipe tobacco, including barring the sales of the products to teens under 18 years old. This is an important step in regulating all products containing tobacco and

  3. #3 peak

    : a minimum age of 16 for a learner's permit; a waiting period of at least six months before

  4. #4 peak

    -fold, according to a study by researchers at RTI International and the Florida Department of Health.⏎The study, published in Pediatrics, found that youth (age 12-17) exposure to electronic cigarette advertisements increased by 256

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. L22:332 suppresses
  2. L13:33 excites
  3. L15:12 weakly excites
  4. L7:229 weakly excites first digit of a split number
  5. L19:258 weakly suppresses
  6. L8:2009 weakly suppresses
  7. L4:448 weakly excites
  8. L3:651 weakly excites zero in dates inside URLs
  9. L22:517 weakly excites
  10. L20:18 weakly excites
  11. L19:2148 weakly excites
  12. L21:1732 weakly suppresses
  13. L11:1852 weakly excites the digit 1 in numbers
  14. L8:20 weakly suppresses
  15. L23:2230 weakly suppresses
  16. L7:8 weakly suppresses
  17. L22:2783 weakly excites
  18. L23:2924 weakly suppresses
  19. L23:2135 weakly excites
  20. L23:2338 weakly suppresses
  21. L24:371 weakly suppresses
  22. L13:2570 weakly excites
  23. L12:1080 weakly excites
  24. L23:600 weakly suppresses

Feeds later layers

  1. L27:132 excites
  2. L27:1196 suppresses
  3. L26:905 suppresses
  4. L27:678 suppresses
  5. L26:580 suppresses
  6. L26:848 weakly excites
  7. L27:205 weakly excites
  8. L27:391 weakly suppresses
  9. L27:763 weakly suppresses
  10. L27:496 weakly suppresses
  11. L27:1202 weakly excites
  12. L27:435 weakly suppresses
  13. L27:153 weakly suppresses the spot just before a historical number
  14. L27:1022 weakly excites
  15. L27:1100 weakly suppresses
  16. L26:2551 weakly excites
  17. L27:1153 weakly excites
  18. L26:2809 weakly excites
  19. L27:992 weakly suppresses
  20. L26:103 weakly suppresses blank space before a number
  21. L27:241 weakly excites
  22. L27:352 weakly suppresses
  23. L27:238 weakly suppresses
  24. L26:667 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 -28.00 and -26.772 values between -25.53 and -24.301 values between -19.36 and -18.131 values between -18.13 and -16.892 values between -15.66 and -14.431 values between -14.43 and -13.192 values between -13.19 and -11.963 values between -11.96 and -10.726 values between -10.72 and -9.496 values between -9.49 and -8.264 values between -8.26 and -7.028 values between -7.02 and -5.796 values between -5.79 and -4.5513 values between -4.55 and -3.3225 values between -3.32 and -2.0951 values between -2.09 and -0.853442 values between -0.85 and 0.38280 values between 0.38 and 1.6219 values between 1.62 and 2.8512 values between 2.85 and 4.085 values between 4.08 and 5.327 values between 5.32 and 6.553 values between 6.55 and 7.791 values between 7.79 and 9.021 values between 9.02 and 10.251 values between 10.25 and 11.491 values between 12.72 and 13.962 values between 13.96 and 15.191 values between 16.42 and 17.661 values between 20.13 and 21.36offnegativepositiveoften

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. ändig
  2. addElement
  3. ioned
  4. Cad
  5. @admin
  6. .outer
  7. fried
  8. eldorf
  9. 蛘
  10. airie

Suppresses

  1. 6
  2. 7
  3. 8
  4. 4
  5. 5
  6. 7
  7. 6
  8. 8
  9. 4
  10. 5