Neuron Explorer

Qwen3-0.6B

slot before a grade or age number

A neuron in layer 23 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 empty spot just before a grade or age number, such as "grades [4]-12" or "ages [8]-12". When it is turned down the model got stuck repeating the word "local". Its opposite side fires on "your" in advice sentences.

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

    students before they get to college. Every year, students submit their great ideas to win prizes and gain the hands-on experience exclusive to our contest!⏎Students in grades 4-12 are invited to compete in this contest honoring innovation & Thomas Edison

  2. #2 peak

    games are significant;⏎They have a tremendous role shaping the child’s life”⏎– e² Young Engineers –⏎Our most popular program introduces elementary school children (grades 1- 5) to the basic principles of mathematics, science and engineering.

  3. #3 peak

    way that is entertaining, accessible and gender-neutral—⏎As the contents of Fizzyx for Fun meet many of the Prescribed Science Learning Outcomes for Grades 6 – 9, it would be an important resource for school libraries and classroom

  4. #4 peak

    hours and half-hours⏎- I know that there are five- minute intervals between each number on the clock face⏎More Lessons for Grade 2⏎Common Core for Grade 2⏎Telling time on the analog and

  5. #5 peak

    White House and Washington, D.C. from 1792 to today.⏎Who was the first U.S. president to be photographed?⏎Grades: 3–5, 6–8⏎A few fun facts about America's

  6. #6 peak

    Kano (Raspberry Pi) computer kit.⏎TWO CAMPS FOR TWO DIFFERENT AGE GROUPS! We have a “Basic” camp for ages 8 – 12 that will create the Kano computer and target exercises and

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: local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local local

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

    are very powerful and can provide you with the insight necessary for any research project.<|endoftext|>Whether it's bringing flowers to a friend feeling under the weather or volunteering at your community soup kitchen, giving to others is a selfless act that benefits everyone involved

  2. #2 peak

    William Penn described the forests of Pennsylvania as “the natural produce of the country.” Consider using native trees in yards to preserve our habitats.⏎For additional assistance, contact your local Penn State Master Gardeners or local Native Plant Society.⏎Return to Pest Pages

  3. #3 peak

    are not guaranteed immunity.⏎All living stateside should prepare for this flu season as if they will come down with the flu.⏎You can get a flu shot at your doctor’s office, but you needn’t make an appointment if it inconven

  4. #4 peak

    been eliminated in Canada; however, imported cases continue to occur.⏎To obtain or inquire about an immunization record, or for general immunization information, please contact your local public health unit.<|endoftext|>states that England officially recognizes the colonies to be in

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. L20:520 weakly excites
  2. L21:2538 weakly suppresses
  3. L19:817 weakly suppresses
  4. L8:636 weakly suppresses
  5. L1:2479 weakly excites
  6. L6:1300 weakly suppresses
  7. L21:919 weakly suppresses
  8. L22:151 weakly suppresses
  9. L7:1043 weakly suppresses
  10. L0:2868 weakly suppresses
  11. L21:1172 weakly excites
  12. L7:1181 weakly excites
  13. L9:143 weakly excites
  14. L2:2170 weakly suppresses
  15. L17:831 weakly suppresses
  16. L0:2572 weakly excites
  17. L4:2880 weakly excites
  18. L7:1353 weakly suppresses
  19. L2:1552 weakly suppresses second half of multi-word names
  20. L3:2207 weakly suppresses
  21. L22:2600 weakly excites
  22. L17:1485 weakly excites
  23. L9:36 weakly suppresses
  24. L14:1434 weakly suppresses

Feeds later layers

  1. L27:1288 strongly suppresses
  2. L25:1832 suppresses the word 'local'
  3. L24:1353 weakly suppresses
  4. L25:1129 weakly suppresses
  5. L24:3036 weakly excites
  6. L24:2767 weakly excites
  7. L24:993 weakly suppresses
  8. L27:1569 weakly excites first pieces of split words like clo and Le
  9. L26:3005 weakly excites
  10. L27:566 weakly suppresses
  11. L24:179 weakly excites
  12. L25:3053 weakly excites
  13. L25:2152 weakly excites
  14. L27:675 weakly suppresses
  15. L25:2333 weakly suppresses
  16. L26:2903 weakly suppresses
  17. L24:2238 weakly suppresses
  18. L25:1015 weakly suppresses
  19. L24:1636 weakly excites
  20. L25:1040 weakly suppresses digit 1 inside dates and times
  21. L26:1900 weakly suppresses
  22. L26:2266 weakly excites
  23. L27:885 weakly suppresses
  24. L24:2017 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 -22.17 and -21.101 values between -21.10 and -20.041 values between -20.04 and -18.973 values between -16.83 and -15.761 values between -15.76 and -14.702 values between -14.70 and -13.632 values between -13.63 and -12.561 values between -12.56 and -11.493 values between -11.49 and -10.425 values between -10.42 and -9.362 values between -9.36 and -8.294 values between -8.29 and -7.2211 values between -7.22 and -6.1511 values between -6.15 and -5.0814 values between -5.08 and -4.0218 values between -4.02 and -2.9533 values between -2.95 and -1.88126 values between -1.88 and -0.812846 values between -0.81 and 0.26761 values between 0.26 and 1.3235 values between 1.32 and 2.3910 values between 2.39 and 3.464 values between 3.46 and 4.532 values between 4.53 and 5.605 values between 5.60 and 6.662 values between 6.66 and 7.732 values between 8.80 and 9.871 values between 14.14 and 15.211 values between 19.48 and 20.55offnegativepositiveoften

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. 斯顿
  2. ersed
  3. SplitOptions
  4. しょう
  5. sess
  6. Himself
  7. _SPLIT
  8. 特斯
  9. culpa
  10. .netflix

Suppresses

  1. local
  2. local
  3. Local
  4. (local
  5. →local
  6. Local
  7. 本地
  8. 当地
  9. -local
  10. 当地的