Neuron Explorer

Qwen3-0.6B

period in 'U.S.' and 'U.K.'

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 fires on the full stop inside abbreviations like "U.S." and "U.K.", where the dot sits in the middle of a country name. When it was turned up, the model's writing did not change in a clear way.

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

    th century. The root word, progress, is defined as moving or “progressing” past historical documents such as the Declration of Independence and the U.S. Constitution. Progressives see these documents as outdated and therefore no longer have relevant meaning

  2. #2 peak

    79, are taking the U.K. government to court over how the country has failed to reach its own targets over climate

  3. #3 peak

    . The U.S. planned to buy nearly 3,000 JSFs for its air force, navy, and Marine Corps, while the U.K. was expected to procure up to 150.⏎At the beginning of the

  4. #4 peak

    we consume<|endoftext|>The first non-stop trans-Atlantic flight⏎THE VERY FIRST NON-STOP TRANS-ATLANTIC FLIGHT⏎Note: The U.S. Navy had already crossed the Atlantic, via the Azores to land in Plymouth previously

  5. #5 peak

    Wednesday.⏎The legal case comes after last week when climate advisors in the U.K. warned that the younger generation will be the ones paying the price of climate change.

  6. #6 peak

    . The city is located on U.S. Route 101 on the shores of Humboldt Bay, north of San

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: "I have been thinking about lately, and I have been thinking about lately. But I have been thinking about lately. But I have been thinking about lately. But I have been thinking about lately. But I have been thinking about lately.

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

    While this may be effective in some cases, it is also ineffective because nicotine is this type of potent substance.⏎Another solution to the problem is to utilize an electronic cigarette that doesn’t contain nicotine at all. By avoiding nicotine, the unit to avoid

  2. #2 peak

    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% from 2011 to

  3. #3 peak

    . Stat. Ann. § 1185 tells us that this cited statute is Section 1185 of Title 23 of the Vermont Statutes Annotated. And the federal citation 42 U.S.C. §

  4. #4 peak

    force behind founder and CEO Melissa Clark-Reynolds’ decision to create MiniMonos in the first place. After Melissa saw the movie ‘An Inconvenient Truth’, she decided to contact ex-US Vice President Al Gore to offer her help.

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. L1:189 weakly suppresses
  2. L0:279 weakly suppresses
  3. L0:1 weakly suppresses
  4. L1:2649 weakly excites
  5. L1:436 weakly suppresses
  6. L1:374 weakly excites
  7. L0:1075 weakly excites
  8. L0:2058 weakly excites
  9. L1:1244 weakly suppresses "of" after "board" or "off"
  10. L1:196 weakly excites
  11. L1:545 weakly suppresses
  12. L0:987 weakly suppresses
  13. L0:2709 weakly excites
  14. L0:2179 weakly suppresses
  15. L1:1583 weakly excites
  16. L1:1159 weakly excites
  17. L0:1425 weakly suppresses
  18. L0:1173 weakly suppresses
  19. L1:388 weakly suppresses
  20. L0:368 weakly suppresses
  21. L0:389 weakly excites
  22. L0:678 weakly suppresses
  23. L1:952 weakly suppresses
  24. L0:2640 weakly suppresses

Feeds later layers

  1. L3:1555 excites
  2. L3:699 excites
  3. L5:2801 weakly suppresses
  4. L4:2514 weakly suppresses
  5. L10:77 weakly excites
  6. L4:68 weakly suppresses
  7. L11:152 weakly suppresses
  8. L12:2916 weakly suppresses
  9. L25:2785 weakly suppresses
  10. L7:1713 weakly excites
  11. L4:1081 weakly suppresses
  12. L23:898 weakly excites
  13. L23:1344 weakly suppresses
  14. L8:775 weakly suppresses
  15. L16:2571 weakly suppresses
  16. L7:148 weakly suppresses
  17. L25:2941 weakly suppresses
  18. L12:1351 weakly excites
  19. L13:2527 weakly suppresses
  20. L15:2121 weakly suppresses
  21. L8:2801 weakly suppresses
  22. L14:298 weakly suppresses
  23. L5:1849 weakly suppresses
  24. L9:119 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.

3 values between -1.08 and -0.813 values between -0.81 and -0.5545 values between -0.55 and -0.281425 values between -0.28 and -0.022334 values between -0.02 and 0.2476 values between 0.24 and 0.518 values between 0.51 and 0.772 values between 0.77 and 1.031 values between 1.03 and 1.304 values between 1.30 and 1.562 values between 1.56 and 1.822 values between 1.82 and 2.091 values between 2.88 and 3.141 values between 8.68 and 8.941 values between 9.21 and 9.47offnegativepositiveoften

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. HIR
  2. 出处
  3. OUR
  4. subsequent
  5. Ặ
  6. led
  7. related
  8. asset
  9. succeeding
  10. aligned

Suppresses

  1. ponsors
  2. iae
  3. WindowState
  4. boards
  5. 窄
  6. endars
  7. ẻ
  8. 向社会
  9. çe
  10. ene