Neuron Explorer

Qwen3-0.6B

the phrase 'by the time'

A neuron in layer 7 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 reacts to the word "time" when it comes after "by the," as in "by the time Nixon reimposed." Turned up, the model wrote more about time and timing, such as "the time was too late." It is slightly negative on country and place names that follow "in" or "Holy," like Korea and Land.

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

    the price — but not until after he’d coasted to a landslide re-election in 1972 over Democratic Sen. George McGovern.⏎By the time Nixon reimposed

  2. #2 peak

    for a Russian prince. News of the child prodigy spread quickly; his first compositions were published in a Warsaw newspaper, and he was famous outside Poland by the time he was 15.⏎Chopin’s first teacher was a man named

  3. #3 peak

    her bicep in the air, saying "...We Can Do It!". Or the the poster warning against spies with "...Loose Lips Sink Ships".⏎By the time you have completed our puzzle, you'll be, once again, very proud of

  4. #4 peak

    a relatively rural area before World War II. There was a large influx of servicemen for training during the war, industry like ship building occurred, and by the time the war was over, we’re getting into what’s considered modern Florida.”⏎According

  5. #5 peak

    66, but it was the oil crisis of 1973/74 that ended the "economic miracle" once and for all. By the time Herbert Grünewald succeeded Kurt Hansen as Management Board Chairman following the 19

  6. #6 peak

    Army's Corps of Topographical Engineers surveying the southwestern and northwestern boundaries of the United States as well the potential routes for a transcontinental railroad. By the time of the Civil War, an ocean-to-ocean American empire with borders clearly defined

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: the best way to manage the risk of a health condition. The primary focus is on the organization that the individual is responsible for. The person who is responsible for the health care coverage, and the individual is responsible for managing the risk of 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

    Nurses, Carolyn Henderson, ORGANIZING THE PROFESSIONS FOR⏎Volunteering For Risk: Black Military Women Overseas during⏎the Wars in Korea and Vietnam - Tells how<|endoftext|>Students know best⏎Who knows best about how your

  2. #2 peak

    additional 670,846 were wounded. They were committed to victory, no matter the personal cost.⏎Since then, our soldiers have fought in Korea, Vietnam, Grenada, Panama, the Balkans, Afghanistan, Iraq and

  3. #3 peak

    which became the seat of the theological faculty of Paris, the most illustrious seat of learning in the medieval period.⏎He led the Seventh Crusade to the Holy Land in 1248–50 and died on

  4. #4 peak

    water.⏎To suggest in any way that bottled water is less stringently regulated than tap water is simply not true...<|endoftext|>By the time the allies landed in Normandy on June 6th, 1944, they had achieved near

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. L6:2332 excites
  2. L6:7 suppresses
  3. L6:3067 weakly excites
  4. L4:2608 weakly suppresses
  5. L5:1431 weakly suppresses
  6. L4:864 weakly excites
  7. L6:1857 weakly excites "until" after a negated clause
  8. L2:621 weakly suppresses
  9. L5:2864 weakly excites 'time' in 'for the first time'
  10. L6:2390 weakly excites
  11. L6:2277 weakly excites
  12. L5:1406 weakly suppresses
  13. L4:1817 weakly suppresses
  14. L0:2731 weakly excites
  15. L5:3046 weakly suppresses
  16. L5:2930 weakly suppresses
  17. L5:2832 weakly suppresses
  18. L4:134 weakly excites
  19. L2:2971 weakly excites
  20. L4:589 weakly suppresses
  21. L5:315 weakly suppresses
  22. L4:314 weakly suppresses
  23. L2:2045 weakly excites
  24. L5:2457 weakly suppresses

Feeds later layers

  1. L11:2600 excites
  2. L8:1808 weakly excites
  3. L9:2391 weakly excites
  4. L11:632 weakly suppresses
  5. L11:1609 weakly suppresses
  6. L10:2712 weakly suppresses
  7. L18:2782 weakly suppresses
  8. L10:1544 weakly suppresses
  9. L9:2795 weakly suppresses
  10. L9:465 weakly excites
  11. L14:1847 weakly suppresses
  12. L22:935 weakly suppresses
  13. L9:2217 weakly excites
  14. L15:728 weakly suppresses
  15. L12:718 weakly excites
  16. L19:3070 weakly excites
  17. L20:309 weakly suppresses
  18. L9:1089 weakly excites
  19. L10:468 weakly excites
  20. L11:239 weakly excites
  21. L17:682 weakly excites
  22. L25:252 weakly suppresses
  23. L14:1662 weakly suppresses
  24. L11:2311 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 -2.21 and -2.111 values between -1.71 and -1.611 values between -1.61 and -1.511 values between -1.21 and -1.111 values between -1.11 and -1.011 values between -1.01 and -0.914 values between -0.91 and -0.813 values between -0.81 and -0.713 values between -0.71 and -0.611 values between -0.61 and -0.5110 values between -0.51 and -0.4151 values between -0.41 and -0.31152 values between -0.31 and -0.21418 values between -0.21 and -0.111043 values between -0.11 and -0.011243 values between -0.01 and 0.09634 values between 0.09 and 0.19219 values between 0.19 and 0.2977 values between 0.29 and 0.3916 values between 0.39 and 0.4910 values between 0.49 and 0.597 values between 0.59 and 0.702 values between 0.70 and 0.802 values between 0.80 and 0.901 values between 0.90 and 1.002 values between 1.00 and 1.101 values between 1.30 and 1.401 values between 1.40 and 1.502 values between 1.70 and 1.80offnegativepositiveoften

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. 빅
  2. later
  3. pack
  4. лин
  5. iese
  6. iles
  7. ikal
  8. oras
  9. افة
  10. mł

Suppresses

  1. aida
  2. Weiner
  3. ito
  4. 瑑
  5. awy
  6. =F
  7. annunci
  8. éro
  9. сог
  10. =R