Neuron Explorer

Qwen3-0.6B

"Great" in proper place names

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 fires on the word "Great" when it is part of a famous name, like "Great Lakes", "Great Plains" or "The Great Gatsby". Turning it down made the model produce endless "of of of" repetition, and turning it up gave loose, unrelated text, so the steering does not clearly reflect 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

    versus broadcasting nutrients is likely to become a much bigger conversation as Ontario works to reduce the impact farm phosphorous on the Great Lakes.⏎In this Corn School episode, Purdue University professor Tony Vyn offers insights

  2. #2 peak

    itzer Dam.⏎- A network of 150 miles of trails, including the Wild Oak National Recreation Trail and a portion of the Great Eastern Trail. The historic Civilian Conservation Corps (CCC) built many of the

  3. #3 peak

    county in Utah by land area and third-smallest by total area.⏎The county extends from high in the Wasatch Range in the east into a portion of the Great Salt Lake to the west. The Weber and Ogden rivers and their tribut

  4. #4 peak

    action.⏎Global Big Day **⏎A tradition among birders, this day challenges bird watchers to record as many species as possible in 24 hours. Like the Great Backyard Bird Count, people can spend as much time as they like, anywhere

  5. #5 peak

    with posters displayed at railway stations, and also small presentation tins containing three or four biscuits that were given free of charge to first class passengers on trains of the Great Western Railway departing from Paddington station, London, with suggestions that they should look

  6. #6 peak

    part decadence, excess, and disillusionment. Being wealthy is certainly one of the main accomplishments that characterized the American society. Through his characters, the novel The Great Gatsby by F. Scott Fitzgerald reveals the consequences of this dream on the population

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: of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of of 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

    Edgar Allen Poe who published The Premature Burial (1844), The Cask of Amontillado (1846) and The Fall of the House of Usher (1839.) They are all good

  2. #2 peak

    on to become a leading member of the National Left Wing Council and editor of that group's Chicago newspaper, The Communist.⏎Following the September 1918 state convention of the Socialist Party of Michigan, a number of delegates, including Keracher

  3. #3 peak

    eenth century evaluations about jesuitic missions among the Guaranis - Relação abbreviada da republica, que os religiosos jesuitas estabelecerão no Paraguai, by Marquis of Pombal,

  4. #4 peak

    What is the Motor Learning Lab?⏎The Motor Learning Lab (MLL) is a new facility in the RNOH therapies department. The MLL features new state-of-the-art equipment for gait analysis (the GRAIL system supplied by Mot

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. L11:2124 weakly excites
  2. L21:2530 weakly suppresses
  3. L6:467 weakly suppresses
  4. L22:344 weakly excites
  5. L21:2671 weakly suppresses
  6. L22:2801 weakly suppresses
  7. L4:2501 weakly excites
  8. L21:2090 weakly suppresses
  9. L15:1949 weakly suppresses
  10. L6:2628 weakly suppresses
  11. L0:2701 weakly excites
  12. L19:402 weakly excites
  13. L11:2885 weakly excites
  14. L11:3009 weakly excites
  15. L13:2595 weakly suppresses the :// after http or https
  16. L4:1314 weakly excites
  17. L0:1625 weakly suppresses
  18. L15:1485 weakly suppresses
  19. L21:2139 weakly excites "of" in "Battle of"
  20. L8:1597 weakly excites
  21. L22:2630 weakly excites
  22. L8:2793 weakly suppresses "life" after origins or search phrases
  23. L5:1969 weakly excites
  24. L14:2839 weakly excites

Feeds later layers

  1. L26:711 suppresses 'years' after an age number
  2. L25:2020 suppresses
  3. L27:447 weakly suppresses
  4. L26:920 weakly suppresses
  5. L24:992 weakly excites
  6. L27:2008 weakly excites
  7. L26:115 weakly excites
  8. L24:860 weakly suppresses
  9. L26:1637 weakly suppresses
  10. L27:1962 weakly suppresses
  11. L27:845 weakly suppresses
  12. L27:813 weakly excites
  13. L24:1772 weakly excites
  14. L27:2289 weakly excites
  15. L25:18 weakly excites
  16. L26:1457 weakly suppresses
  17. L24:962 weakly suppresses
  18. L26:1412 weakly excites
  19. L24:1671 weakly suppresses
  20. L24:556 weakly suppresses
  21. L25:1107 weakly suppresses
  22. L27:2964 weakly suppresses
  23. L24:2753 weakly excites
  24. L27:1976 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 -25.33 and -24.261 values between -16.75 and -15.671 values between -13.53 and -12.461 values between -11.38 and -10.311 values between -10.31 and -9.242 values between -9.24 and -8.179 values between -8.17 and -7.0913 values between -7.09 and -6.029 values between -6.02 and -4.9530 values between -4.95 and -3.8842 values between -3.88 and -2.80100 values between -2.80 and -1.73260 values between -1.73 and -0.662515 values between -0.66 and 0.42540 values between 0.42 and 1.49143 values between 1.49 and 2.5695 values between 2.56 and 3.6344 values between 3.63 and 4.7142 values between 4.71 and 5.7820 values between 5.78 and 6.858 values between 6.85 and 7.928 values between 7.92 and 9.008 values between 9.00 and 10.076 values between 10.07 and 11.143 values between 11.14 and 12.222 values between 12.22 and 13.292 values between 13.29 and 14.361 values between 14.36 and 15.431 values between 16.51 and 17.58offnegativepositiveoften

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. 她是
  2. 她
  3. 转移
  4. her
  5. She
  6. she
  7. 她的
  8. 发展的
  9. 转变
  10. 穿

Suppresses

  1. of
  2. →of
  3. -of
  4. Of
  5. of
  6. acus
  7. _OF
  8. _of
  9. ofs
  10. ばかり