Neuron Explorer

Qwen3-0.6B

unclear: names and 're' word starts

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

The pattern is unclear. It fires on some names, such as Yuri Gagarin, and on 'Re' or 're' at the start of words like 'reimagining' and 'rely', but these examples do not share a clear theme. Turning it up gave unrelated text about numbers and meetings.

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

    6, vol.25, n.1, pp. 44-69. ISSN 1980-4369. http://dx.doi.org/10.1590/S01

  2. #2 peak

    American History: Selected Documents Celebrating the Manuscript Division’s First 100 Years features drafts of Langston Hughes’ poem, “The Ballad of Booker T.” In this 1941 poem,<|endoftext|>Whenever the stench

  3. #3 peak

    11 was a victory for the U.S. in the Cold War Space Race with the Soviet Union who, in 1961, had sent Yuri Gagarin to the Cosmos to become the first human in outer space and the

  4. #4 peak

    2014, Whales as marine ecosystem engineers, Front Ecol Environ, http://www.joeroman.com/wordpress/wp-content/uploads/201

  5. #5 peak

    WHO’s 13th General Program of Work target of ‘1 billion lives made healthier’ by 2023 and the global Education 2030 Agenda coordinated by UNESCO.⏎The global standards provide a resource for education systems to

  6. #6 peak

    told Sputnik.⏎“We will always remember the names of the first daredevils who made the first flight into space and the first round the world trip – Yuri Gagarin and Ferdinand Magellan. Without remembering the past, we will not

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 just found a new kind of cappardard on the market. It looks just like a cappardard but harder, harder, harder hard, hard, hard hard hard hard hard hard hard hard hard hard hard hard hard hard

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

    other artistic production where sign language can be deeply explore and fully understand. Ryan Commerson, producer and host of the movie “Media, Power & Ideology: Re-Presenting D-E-A-D”, states that all the preconceptions societies

  2. #2 peak

    - 45 drops<|endoftext|>BROOKLYN BRIDGE FOREST⏎Our Brooklyn Bridge Forest submission was the winner of Van Alen Institute’s competition ‘Reimagining Brooklyn Bridge’, ‘which challenged participants to rethink the iconic Brooklyn Bridge walk

  3. #3 peak

    top of the bee habitat is garden sage. The Lehigh Valley’s many pollinators—hummingbirds, bees and butterflies, to name a few—rely on these plants for precious nutrients.⏎These gardens are small and easy to maintain

  4. #4 peak

    few years later<|endoftext|>Hand-coloured lithograph entitled; 'Dampfmaschinen Pferd worauf man in einer stunde von Paris nach Petersburg reiten kann' ('Steam engine horse on which one can

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. L17:2796 suppresses
  2. L4:101 excites
  3. L3:515 weakly excites
  4. L15:1659 weakly suppresses
  5. L23:1143 weakly excites
  6. L2:2773 weakly excites
  7. L5:543 weakly excites
  8. L22:2239 weakly suppresses
  9. L22:148 weakly suppresses
  10. L23:903 weakly excites
  11. L20:1387 weakly excites
  12. L16:390 weakly excites
  13. L13:262 weakly suppresses
  14. L11:1997 weakly suppresses
  15. L18:2313 weakly excites
  16. L15:618 weakly suppresses
  17. L5:1952 weakly excites
  18. L11:2785 weakly suppresses
  19. L18:2392 weakly excites
  20. L19:1131 weakly excites
  21. L7:227 weakly excites word fragments inside split words
  22. L2:2738 weakly suppresses
  23. L1:76 weakly excites
  24. L17:478 weakly suppresses

Feeds later layers

  1. L26:1055 weakly suppresses
  2. L27:1721 weakly suppresses
  3. L27:1560 weakly suppresses
  4. L26:2912 weakly excites
  5. L27:918 weakly excites
  6. L26:370 weakly excites
  7. L26:1482 weakly excites
  8. L26:786 weakly excites
  9. L27:2830 weakly excites
  10. L27:1558 weakly excites
  11. L27:2679 weakly suppresses
  12. L26:1145 weakly suppresses
  13. L27:2496 weakly suppresses
  14. L27:1085 weakly suppresses
  15. L26:2611 weakly excites
  16. L26:2087 weakly suppresses
  17. L26:2248 weakly suppresses
  18. L26:1902 weakly suppresses
  19. L26:2342 weakly excites
  20. L26:2095 weakly suppresses
  21. L27:1400 weakly suppresses
  22. L27:229 weakly suppresses
  23. L27:1154 weakly suppresses
  24. L26:2323 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 -14.16 and -13.531 values between -13.53 and -12.912 values between -11.04 and -10.421 values between -10.42 and -9.791 values between -9.17 and -8.551 values between -7.92 and -7.301 values between -7.30 and -6.686 values between -6.68 and -6.052 values between -6.05 and -5.435 values between -5.43 and -4.8111 values between -4.81 and -4.1810 values between -4.18 and -3.5611 values between -3.56 and -2.9315 values between -2.93 and -2.3130 values between -2.31 and -1.6969 values between -1.69 and -1.06545 values between -1.06 and -0.442348 values between -0.44 and 0.18575 values between 0.18 and 0.81110 values between 0.81 and 1.4359 values between 1.43 and 2.0539 values between 2.05 and 2.6816 values between 2.68 and 3.3018 values between 3.30 and 3.9211 values between 3.92 and 4.555 values between 4.55 and 5.172 values between 5.17 and 5.793 values between 5.79 and 6.424 values between 6.42 and 7.041 values between 7.04 and 7.661 values between 8.29 and 8.911 values between 8.91 and 9.531 values between 9.53 and 10.162 values between 10.16 and 10.78offnegativepositiveoften

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. ità
  2. ilar
  3. 晚
  4. ress
  5. atures
  6. ýt
  7. gage
  8. ymi
  9. vasion
  10. adamente

Suppresses

  1. IService
  2. 破
  3. kind
  4. breaking
  5. 积蓄
  6. anga
  7. <th
  8. едь
  9. 服务
  10. Service