Neuron Explorer

Qwen3-0.6B

'time' in 'for the first time'

A neuron in layer 2 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 responds to the word 'time' in the phrase 'for the first time'. Its opposite side is the ending 'hs' in spelled-out fractions like 'sixteenths' or 'ninths'. Turning it up gave text that wandered into unrelated topics, so the steering doesn't show the trigger clearly.

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

    to verify proper control and operation can ensure longer system life<|endoftext|>Asking the Right Questions About Your Child's Camp⏎Sending your child to summer camp for the first time can be a difficult experience; from asking about camp size to camp cafeteria food,

  2. #2 peak

    to bring in the new school year.⏎Keeping children safe is the Olympia School District’s top priority, especially for younger students and those heading to school for the first time. Parents and/or teachers can take the following steps:⏎- Teach students not to

  3. #3 peak

    2012 - 18:30 in Astronomy & Space⏎NASA's rover Curiosity touched a Martian rock with its robotic arm for the first time on Sept. 22, assessing what chemical elements are in the rock called

  4. #4 peak

    only half the amount added in 2018. Energy storage – previously “on track” – has been downgraded as annual installations fell for the first time in nearly a decade.⏎Many of the largest sources of global emissions remain “off

  5. #5 peak

    repeated it a few times since then.⏎But there were two new dimensions of the recent flight. One, it carried a Chinese woman astronaut into space for the first time. Secondly and far more significantly, it demonstrated China's capability to conduct docking of

  6. #6 peak

    a mistake has never tried anything new.” ~ Albert Einstein⏎When we’re small children, our mistakes are applauded. Our falling-down attempts to walk for the first time are cheered by our parents, giving us the courage to get up and try again

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: 12345678910112345678910112345678910112345678910

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

    1/4 = 4/16 and three sixteenths is 3/16, so one quarter is greater than three sixteenths⏎no, 3/16 is less than 1/2⏎

  2. #2 peak

    to one third. One third of anything is a greater quantity than one quarter of anything, therefore, "Three-ninths" is the larger of the two.⏎One fourth is a quarter and three fourth

  3. #3 peak

    0. Compare that with the seaboard of France and England. What sort of coast is it? There were shoals and sand everywhere, blocking nine-tenths of the land altogether, and doing their best to block the other tenth where the

  4. #4 peak

    in cotton buds in such areas.⏎The boll weevil is a small beetle, measuring only about two-tenths of an inch long, commonly known to feed and lay eggs in cotton buds and

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. L1:1926 weakly suppresses
  2. L0:950 weakly suppresses
  3. L1:2229 weakly excites
  4. L1:1162 weakly suppresses
  5. L0:2559 weakly excites
  6. L1:1207 weakly excites unclear: prepositions after nouns
  7. L0:532 weakly suppresses
  8. L1:1231 weakly excites
  9. L1:137 weakly excites
  10. L1:169 weakly excites
  11. L0:1300 weakly excites
  12. L0:444 weakly excites
  13. L1:382 weakly suppresses
  14. L0:2107 weakly suppresses
  15. L0:1795 weakly excites
  16. L0:1184 weakly excites
  17. L1:1803 weakly excites
  18. L0:2817 weakly suppresses
  19. L0:2110 weakly suppresses
  20. L1:463 weakly excites
  21. L0:1507 weakly suppresses
  22. L1:2924 weakly suppresses
  23. L1:2960 weakly suppresses
  24. L1:2594 weakly suppresses

Feeds later layers

  1. L4:448 suppresses
  2. L10:1988 weakly suppresses
  3. L6:1147 weakly excites
  4. L7:2648 weakly suppresses
  5. L3:1129 weakly suppresses
  6. L4:846 weakly suppresses
  7. L15:711 weakly suppresses
  8. L27:1213 weakly suppresses
  9. L9:1241 weakly suppresses
  10. L16:724 weakly suppresses
  11. L20:2970 weakly excites
  12. L8:2138 weakly suppresses unclear: mid-name word pieces
  13. L12:752 weakly suppresses
  14. L4:2364 weakly suppresses
  15. L22:2283 weakly suppresses
  16. L8:41 weakly suppresses
  17. L7:1607 weakly suppresses
  18. L10:2806 weakly excites
  19. L23:1175 weakly excites
  20. L10:1518 weakly suppresses
  21. L12:510 weakly suppresses
  22. L16:1746 weakly excites
  23. L6:323 weakly suppresses
  24. L22:664 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 -2.54 and -2.441 values between -1.22 and -1.131 values between -0.94 and -0.851 values between -0.85 and -0.751 values between -0.75 and -0.662 values between -0.66 and -0.572 values between -0.57 and -0.4710 values between -0.47 and -0.3824 values between -0.38 and -0.28100 values between -0.28 and -0.19411 values between -0.19 and -0.101394 values between -0.10 and -0.001474 values between -0.00 and 0.09368 values between 0.09 and 0.1971 values between 0.19 and 0.2819 values between 0.28 and 0.3713 values between 0.37 and 0.476 values between 0.47 and 0.563 values between 0.56 and 0.662 values between 0.66 and 0.751 values between 0.75 and 0.841 values between 0.94 and 1.031 values between 1.03 and 1.121 values between 1.12 and 1.22offnegativepositiveoften

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. caption
  2. sterol
  3. isen
  4. summary
  5. 强者
  6. تكامل
  7. stown
  8. issen
  9. atten
  10. agues

Suppresses

  1. ICLE
  2. iology
  3. Elvis
  4. ağı
  5. eldo
  6. 仗
  7. ANNEL
  8. olare
  9. ое
  10. pró