Neuron Explorer

Qwen3-0.6B

mixed: 'New' in place names, square roots

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 is mixed. It mainly reacts to the word 'New' in place names like New York or New Mexico, and also to square-root signs in maths. Turning it up or down produced repetitive text about 'has' and 'have', which matches the words it tends to suppress.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    limit it gives you the function being integrated.⏎Thus (because the lower limit doesn't play any role being a constant number)⏎dF(x)/dx2 = √(1 + t3) where t = x2 (2)⏎If

  2. #2 peak

    the speeds of two satellites placed in stable circular orbits of radius R and 2R respectively?⏎C) 1 / 2⏎D) 1 / √2⏎E) 2 √2⏎What is the centrip

  3. #3 peak

    lenker, a Senior Researcher at Institut Jean-Nicod within France's National Center for Scientific Research (CNRS) and a Global Distinguished Professor at New York University.⏎The combined team of linguists and primatologists analyzed alarm calls

  4. #4 peak

    Biology at the University of New Mexico. Troy L. Best is a professor of biological sciences at Auburn University.

  5. #5 peak

    tutorial from the Australia Telescope National Facility, Parkes, New South Wales⏎Amateurfunk, Erde-Mond-Erde,

  6. #6 peak

    detection being between 60 and 70 years.⏎“Cancer is a disease of the aging,” said Dr. Deborah Axelrod, associate professor at New York University (NYU) Langone Medical Center’s Division of Breast Surgery.

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: having a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has a has 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

    Edison (1847-1931) in 1878. Edison redesigned the telephone shortly after its invention by Alexander Graham Bell (1847-1922) in 1876.

  2. #2 peak

    <|endoftext|>West Virginia Population 2017⏎Located in the central eastern sector of the United States of America, West Virginia was admitted into the Union in 1863. It is the 41st largest and the 3

  3. #3 peak

    and wear appropriate clothing.⏎Suspects can invoke the Sixth Amendment right to counsel during lineups and showups held after they have been arraigned or indicted. They have the right to have an attorney present, but the attorney only participates as an observer

  4. #4 peak

    and quitting any form of tobacco use are ways to prevent or delay the onset of type 2 diabetes. In a healthier world, everyone, regardless of socioeconomic status, would have access to sustainable, affordable and healthier diets

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. L21:162 excites
  2. L21:2804 excites
  3. L22:2524 excites
  4. L21:2212 suppresses
  5. L22:2752 suppresses
  6. L20:2854 suppresses 'official' and government institution words
  7. L19:2900 excites
  8. L14:1884 excites
  9. L15:559 excites
  10. L19:1959 weakly suppresses
  11. L17:2930 weakly excites
  12. L21:2448 weakly suppresses "a" after give or get verbs
  13. L22:1402 weakly suppresses
  14. L22:2775 weakly excites
  15. L16:1367 weakly excites
  16. L19:1371 weakly suppresses
  17. L18:2198 weakly suppresses
  18. L12:1136 weakly suppresses
  19. L22:3060 weakly excites
  20. L19:377 weakly suppresses
  21. L11:374 weakly excites
  22. L23:387 weakly suppresses
  23. L9:2950 weakly suppresses
  24. L19:1461 weakly suppresses

Feeds later layers

  1. L27:756 strongly suppresses
  2. L26:2817 suppresses
  3. L25:613 excites
  4. L25:685 suppresses
  5. L25:608 weakly excites
  6. L27:1095 weakly suppresses
  7. L27:564 weakly excites
  8. L26:980 weakly excites
  9. L25:929 weakly suppresses
  10. L25:1671 weakly excites
  11. L26:2563 weakly suppresses
  12. L26:2875 weakly excites
  13. L25:1873 weakly suppresses
  14. L25:2702 weakly excites
  15. L27:2364 weakly excites
  16. L25:135 weakly excites
  17. L27:419 weakly excites
  18. L27:2469 weakly excites
  19. L25:1623 weakly excites
  20. L26:2349 weakly excites
  21. L26:1914 weakly suppresses
  22. L26:2462 weakly excites
  23. L26:648 weakly suppresses
  24. L25:598 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 -27.38 and -26.241 values between -21.68 and -20.541 values between -18.27 and -17.132 values between -12.57 and -11.433 values between -10.30 and -9.163 values between -6.88 and -5.742 values between -5.74 and -4.604 values between -4.60 and -3.4616 values between -3.46 and -2.3233 values between -2.32 and -1.191115 values between -1.19 and -0.052604 values between -0.05 and 1.09102 values between 1.09 and 2.239 values between 2.23 and 3.372 values between 3.37 and 4.512 values between 4.51 and 5.651 values between 5.65 and 6.791 values between 6.79 and 7.923 values between 7.92 and 9.061 values between 14.76 and 15.901 values between 15.90 and 17.031 values between 17.03 and 18.17offnegativepositiveoften

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. inely
  2. itti
  3. orsi
  4. idia
  5. ibili
  6. nah
  7. станавли
  8. inaire
  9. outines
  10. mitted

Suppresses

  1. 有
  2. has
  3. have
  4. has
  5. (has
  6. .has
  7. having
  8. Have
  9. had
  10. have