Neuron Explorer

Qwen3-0.6B

unit conversions in parentheses

A neuron in layer 24 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 opening bracket that follows a measurement, such as '6,962 m (22,841 ft)' or '9 inches (23 centimeters)'. It is the bracket that starts a conversion into other units. Turning it up made the writing break into number-heavy and unclear text, so the steering shows nothing about this pattern.

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

    are the highest mountains of each of the seven continents.⏎Aconcagua, South America⏎With its cone-shaped peak rising at 6,962 m (22,841 ft), Cerro Aconcagua holds the title

  2. #2 peak

    the image has a resolution of about 70 meters (230 feet) per pixel. This image was acquired during the HAM

  3. #3 peak

    9.6 km/h)|⏎|Range:||3,500 nautical miles (6,482 km) at 12 knots (22.2

  4. #4 peak

    Visayan warty pigs seasonally grow large manes that start at the top of their head and travel partially down their back and grow up to 9 inches (23 centimeters) long! The mama of the group is Kurit (

  5. #5 peak

    |Dimensions||H: 11 1/16 in. (28.2 cm.), W: 8 7/8 in. (22.5 cm.)|⏎|Location||Alte Pinakothek

  6. #6 peak

    990|⏎|-||Restored||August 21, 1991|⏎|-||Total||64,589 km2 (124th)⏎24,938 sq mi⏎|-

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: "11111111111111111111111111111111111111111111111

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

    40), the first of three volumes on the selected writings of respected English Romantics Anna Laetitia Barbauld and her brother, John Aikin, Sarah Hale explains that the

  2. #2 peak

    of carbon should be kept low for industrial and machinery lubrication.⏎* Malaysian Palm Oil Board (MPOB),⏎6 Persiaran Institusi, Bandar Baru Bangi,⏎43000 Kajang, Selangor

  3. #3 peak

    Generally the ticks that cause Lyme Disease in the USA carry the Borrelia Burgdorferi bacteria in their gut, whilst in Europe it is often the Borrelia Afzelii; these bacteria cause Lyme disease. Currently research

  4. #4 peak

    attempt by society members in 1897 to impose on her an alternative dress code to the customary wig and gown, her exclusion from the society's annual Bar dinners, and the<|endoftext|>In Year 4, our class text this half term

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. L22:1795 suppresses
  2. L22:2327 excites
  3. L21:879 suppresses
  4. L22:1573 excites
  5. L23:2720 excites
  6. L22:1190 excites
  7. L22:854 excites
  8. L19:2609 suppresses
  9. L22:1432 excites
  10. L21:1378 suppresses
  11. L20:256 excites
  12. L18:385 weakly excites
  13. L23:554 weakly suppresses
  14. L23:2728 weakly excites second digit of years like 2020
  15. L10:2719 weakly suppresses
  16. L23:808 weakly suppresses
  17. L21:2586 weakly excites
  18. L23:2035 weakly excites
  19. L17:2227 weakly suppresses
  20. L23:2760 weakly suppresses
  21. L22:176 weakly suppresses
  22. L23:1272 weakly excites
  23. L23:723 weakly excites metric or imperial conversions in brackets
  24. L23:399 weakly suppresses

Feeds later layers

  1. L27:189 suppresses
  2. L26:270 suppresses
  3. L26:834 excites
  4. L27:352 excites
  5. L25:1749 suppresses
  6. L25:379 excites
  7. L26:463 suppresses digits inside years and numbers
  8. L27:678 weakly suppresses
  9. L25:2298 weakly excites
  10. L27:364 weakly suppresses
  11. L25:2519 weakly suppresses
  12. L25:1452 weakly suppresses
  13. L26:598 weakly suppresses spot just before a year or century
  14. L25:492 weakly suppresses
  15. L25:145 weakly suppresses
  16. L27:992 weakly excites
  17. L27:2706 weakly suppresses
  18. L25:1287 weakly suppresses
  19. L26:1734 weakly suppresses 'the/as' before 'following'
  20. L25:621 weakly suppresses
  21. L26:2648 weakly suppresses
  22. L25:1322 weakly suppresses
  23. L25:1606 weakly suppresses
  24. L26:556 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 -27.84 and -26.881 values between -17.22 and -16.251 values between -12.39 and -11.421 values between -9.49 and -8.521 values between -8.52 and -7.561 values between -7.56 and -6.598 values between -6.59 and -5.636 values between -5.63 and -4.663 values between -4.66 and -3.6914 values between -3.69 and -2.7313 values between -2.73 and -1.7656 values between -1.76 and -0.802480 values between -0.80 and 0.171262 values between 0.17 and 1.1428 values between 1.14 and 2.1010 values between 2.10 and 3.077 values between 3.07 and 4.046 values between 4.04 and 5.003 values between 5.00 and 5.975 values between 5.97 and 6.931 values between 9.83 and 10.80offnegativepositiveoften

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. yscale
  2. lığın
  3. علامات
  4. marked
  5. ики
  6. eworthy
  7. :+:
  8. aller
  9. Microwave
  10. 维奇

Suppresses

  1. bell
  2. LEE
  3. ront
  4. .Sp
  5. incorpor
  6. PERT
  7. .SP
  8. 涌现
  9. oreal
  10. son