Neuron Explorer

Qwen3-0.6B

the 'hum' piece of words

A neuron in layer 23 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 lights up on the letters 'hum' when they start a word like 'humankind', 'humbly' or 'humongous'. Turning it up made the model drift into repeated, loose phrases about relationships and people, which does not clearly tie back to the trigger.

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

    basis to abolish slavery and assure equal political rights for black Americans. As the fight goes on to assert and realize the Constitutional promise of equal representation for all, we humbly honor and celebrate the lives lost in the ongoing battle toward equality for black

  2. #2 peak

    lost its credibility”. So, today it is said there is no unified story of humankind. There is no grand narrative or what some term metanarrative.⏎

  3. #3 peak

    themselves up to their manners and population. Those living in small groups do not need the same prefect organisation as the very big populations. This thesis is valid also for humankind in small and big developed sites around the world. Capitalists and technocrats

  4. #4 peak

    any hope of avoiding catastrophic climate change means limiting CO2 concentrations below 450 parts per million by 2100. Alternative energies will only get humankind so far, and a vast removal of CO2 will be an important tool

  5. #5 peak

    to the depth of 600 and 700 feet below the sea which would mean stacking up two Statues of Liberty one over another.⏎The humungous piece of ice floating in the Antarctic has a weight of 1 trillion

  6. #6 peak

    seven million years ago. We are the only leaf remaining on our twig, and we are dependent on the same tree we share with millions of other leaves…⏎As humankind, we are all part of the same leaf…⏎The differences we have seen

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: What is the value of $ \overover{22222222222222222222222222222222222222

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

    and contemporaries, including the Roman emperor Caesar Augustus, queen Cleopatra of Egypt and the Nabataean king Obodas III, issued coins emblazoned with their portraits.⏎But as the ruler of a Jewish kingdom where the

  2. #2 peak

    in a flash if you don't have a method for attacking it.<|endoftext|>304 North Cardinal St.⏎Dorchester Center, MA 02124⏎When you’re manually operating a boat, the act of operating it is

  3. #3 peak

    . Secondly, adhering to coding best practices such as modular and reusable code, efficient algorithms, and proper documentation enhances the app’s maintainability and scalability. Embracing agile development methodologies and collaboration tools promotes effective teamwork and streamlines the development process

  4. #4 peak

    They were perceived as “marginal members of the human race,” were denied the right to become naturalized U.S. citizens, and segregated to ethnic enclaves resulting in the rise of Chinatowns which were a

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. L19:334 suppresses
  2. L21:1233 suppresses
  3. L11:2579 weakly suppresses
  4. L2:1923 weakly suppresses
  5. L17:391 weakly excites
  6. L16:55 weakly excites
  7. L20:1504 weakly excites
  8. L2:2278 weakly excites
  9. L0:169 weakly excites
  10. L21:2235 weakly suppresses
  11. L21:2981 weakly suppresses
  12. L1:27 weakly excites
  13. L21:237 weakly suppresses
  14. L13:792 weakly suppresses
  15. L5:552 weakly suppresses the word 'below' pointing to content further down
  16. L13:1965 weakly excites
  17. L13:486 weakly excites
  18. L20:2203 weakly suppresses
  19. L9:593 weakly suppresses
  20. L18:282 weakly excites
  21. L6:284 weakly suppresses
  22. L15:1608 weakly excites
  23. L7:2412 weakly suppresses
  24. L4:466 weakly suppresses

Feeds later layers

  1. L26:1908 weakly excites
  2. L25:1846 weakly excites
  3. L26:1736 weakly excites
  4. L25:2348 weakly excites
  5. L24:2717 weakly excites
  6. L24:2542 weakly suppresses
  7. L24:283 weakly suppresses
  8. L24:72 weakly excites
  9. L27:1326 weakly excites
  10. L25:2229 weakly excites
  11. L24:2317 weakly excites
  12. L24:1252 weakly suppresses
  13. L24:1682 weakly excites
  14. L24:741 weakly excites
  15. L26:2799 weakly suppresses
  16. L24:772 weakly excites
  17. L26:1811 weakly suppresses
  18. L27:2497 weakly suppresses
  19. L24:954 weakly excites
  20. L25:359 weakly suppresses
  21. L26:2332 weakly excites
  22. L24:800 weakly suppresses colon in clock times
  23. L25:2115 weakly excites
  24. L27:1179 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 -20.23 and -18.761 values between -17.28 and -15.801 values between -15.80 and -14.322 values between -11.36 and -9.882 values between -9.88 and -8.402 values between -8.40 and -6.922 values between -6.92 and -5.443 values between -5.44 and -3.966 values between -3.96 and -2.4827 values between -2.48 and -1.003716 values between -1.00 and 0.48119 values between 0.48 and 1.968 values between 1.96 and 3.433 values between 3.43 and 4.914 values between 4.91 and 6.391 values between 6.39 and 7.872 values between 7.87 and 9.352 values between 9.35 and 10.832 values between 10.83 and 12.312 values between 12.31 and 13.791 values between 13.79 and 15.271 values between 37.46 and 38.94offnegativepositiveoften

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. rence
  2. orrar
  3. rise
  4. érie
  5. 釐
  6. itsu
  7. ork
  8. forwards
  9. zier
  10. lane

Suppresses

  1. oha
  2. hele
  3. Velvet
  4. 势
  5. halten
  6. igan
  7. geme
  8. pecial
  9. pecified
  10. mpl