Neuron Explorer

Qwen3-0.6B

unclear: word before an event or name

A neuron in layer 24 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

The snippets are mixed: it fires on words in history-style sentences, like 'in' after 'died' or 'World' before 'War', where a name or event often comes next. Turning it up made the model output the single word 'a' over and over.

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

    romatography<|endoftext|>President Manuel Zelaya and his opponents now in charge in Honduras remain in a standoff. Inside the country, supporters of both sides are waging mass protests, while concerns continue regarding media censorship. This crisis provides an opportunity to

  2. #2 peak

    the leading figure of the artistic school called Neo-Classicism. Born in Paris, he was deserted by his mother at age 9 after his father had been killed in a duel. Raised by uncles, he enrolled at age seventeen in the Royal

  3. #3 peak

    Indian groups throughout the Pacific Northwest, including the Nez Perce, Umatilla and Yakama.⏎Within months, hostilities broke out, now known as the Battle of Frenchtown (also known as the Battle of Wall

  4. #4 peak

    🇱 It is worth noting that the Indian contribution to victory in the First World War in the Middle Eastern area was also noted on one of the 4 stamps

  5. #5 peak

    Palm Oil Board (MPOB),⏎6 Persiaran Institusi, Bandar Baru Bangi,⏎43000 Kajang, Selangor, Malaysia.<|endoftext|>Available for download after purchase.⏎This book is designed to

  6. #6 peak

    example sentence"⏎- S: (n) fire (the event of something burning (often destructive)) "they lost everything in the fire"⏎- S: (n) fire, firing (the act of

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: the way the Misle sold the information to the market, the way the Harks sold the information to the market, and the way the Misle Harks sold the information to the market. Could this be the basis for the basis of

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

    .<|endoftext|>Published 4th February 2021Pre-order⏎In 1863, the US Civil War was raging. Abraham Lincoln’s Emancipation Proclamation had freed over

  2. #2 peak

    of freedom, protection and pay.⏎Additional Details from our Exhibits and Publications:⏎“And retaliation will be our practice now . . .”⏎After President Lincoln issued the Emancipation Proclamation on January 1, 1863, the

  3. #3 peak

    Ice Age that took place long ago.⏎Such an Ice Age is foretold to take place again right before the final battle:"Then will come the second Fimbulwinter. Snow will drive in from all directions; the cold will be severe

  4. #4 peak

    we were introduced to the primal ice first, and so it is when the worlds are to be destroyed, at least in a sense. This is the second Fimbulwinter, where the elements themselves begin to turn against the divine order and rip

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. L22:2847 excites
  2. L21:2240 suppresses
  3. L22:2304 suppresses
  4. L20:2683 excites
  5. L5:2685 weakly excites
  6. L23:1782 weakly suppresses
  7. L22:2276 weakly suppresses
  8. L22:280 weakly suppresses
  9. L18:2063 weakly suppresses
  10. L16:1836 weakly excites
  11. L21:87 weakly excites
  12. L12:1625 weakly suppresses
  13. L3:736 weakly suppresses
  14. L23:2331 weakly excites
  15. L15:2353 weakly excites
  16. L21:1221 weakly suppresses
  17. L6:2808 weakly suppresses
  18. L22:2559 weakly suppresses
  19. L17:593 weakly excites
  20. L1:2102 weakly suppresses
  21. L3:1695 weakly excites opening bracket before a citation
  22. L2:588 weakly excites
  23. L15:889 weakly suppresses end of text after an abrupt cut
  24. L8:1330 weakly excites

Feeds later layers

  1. L27:231 suppresses
  2. L25:1586 excites
  3. L27:797 excites
  4. L27:709 excites
  5. L27:663 excites
  6. L26:1102 suppresses fibromyalgia and hour counts
  7. L25:1861 excites
  8. L25:805 weakly excites
  9. L27:288 weakly excites
  10. L26:829 weakly excites
  11. L25:45 weakly suppresses
  12. L26:484 weakly excites
  13. L26:213 weakly suppresses
  14. L26:546 weakly excites
  15. L26:411 weakly suppresses
  16. L25:1132 weakly suppresses
  17. L25:1314 weakly excites philosophy and ethics terms
  18. L25:627 weakly suppresses
  19. L26:1517 weakly suppresses
  20. L27:173 weakly suppresses
  21. L27:2775 weakly suppresses
  22. L26:2497 weakly suppresses
  23. L25:2924 weakly suppresses
  24. L25:544 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 -18.03 and -17.231 values between -11.60 and -10.791 values between -10.79 and -9.992 values between -9.99 and -9.182 values between -7.58 and -6.771 values between -5.97 and -5.162 values between -5.16 and -4.361 values between -4.36 and -3.555 values between -3.55 and -2.758 values between -2.75 and -1.9519 values between -1.95 and -1.14830 values between -1.14 and -0.342790 values between -0.34 and 0.47153 values between 0.47 and 1.2731 values between 1.27 and 2.0820 values between 2.08 and 2.8812 values between 2.88 and 3.6911 values between 3.69 and 4.494 values between 4.49 and 5.291 values between 5.29 and 6.103 values between 6.10 and 6.904 values between 6.90 and 7.711 values between 7.71 and 8.512 values between 8.51 and 9.321 values between 10.12 and 10.921 values between 10.92 and 11.731 values between 13.34 and 14.14offnegativepositiveoften

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. a
  2. a
  3. levance
  4. getVar
  5. ßer
  6. 颀
  7. adiator
  8. 这是一个
  9. 是一个
  10. esi

Suppresses

  1. arks
  2. sell
  3. ièrement
  4. pedest
  5. breakpoint
  6. ayas
  7. ords
  8. /cms
  9. Grain
  10. _bases