Neuron Explorer

Qwen3-0.6B

hyphen joining two word parts

A neuron in layer 26 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 a hyphen that glues two pieces of a word together, such as 'Bed-Ins' or 'co-creators'. Turning it up made the model's text drift into odd, repetitive loops, so the steering results say little about its role.

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

    o Arama'ah: Tabulae in grammaticen linguae Chaldaeae (Paris, 1560); Aseret ha-Devarim: Decalogus, with the commentary of Abraham Ibn Ezra,

  2. #2 peak

    oko Ono's youth to the violent death of Lennon's mother during his adolescence, from Lennon's post-Ed Sullivan skepticism to John and Yoko's Bed-Ins of 1969 - and unearths the secrets of its

  3. #3 peak

    a curb would be difficult, if not impossible to do without the ability to maintain single limb support for a given amount of time. The ability to switch from two- to one-leg standing is required to perform turns, climb stairs and dress.⏎As

  4. #4 peak

    . In case you need a good example here is one from Edutopia.http://www.edutopia.org/blog/flip-stem-classroom-ainissa-ramirez and all flipped classrooms do not look alike.⏎To see how technology transforms

  5. #5 peak

    of women and men to sustain the world. But it explicitly assigns women the role as sustainers of the world, again with the implicit understanding that men are co-creators and co-sustainers with her. The central source of the ethical

  6. #6 peak

    MIT News Office. 1992-04-15. http://web.mit.edu/newsoffice/1992/citation-0415.html. Retrieved 200<|endoftext|>Our herd of awesome

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: obviously, obviously, obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously

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

    The US federal tax on gasoline has remained at 18.4 cents per gallon during this period of rapid growth in gasoline prices, while state

  2. #2 peak

    the Belgrade Charter and the Tbilisi Declaration can be found at gdrc.org/uem/ee/belgrade.html and gdrc.org/uem/ee/tbilisi.html.⏎KAEE is one of the country’s oldest associations

  3. #3 peak

    undefined constant user_level - assumed 'user_level' (this will throw an Error in a future version of PHP) in /home/commons/public_html/wp-content/plugins/ultimate-google-analytics/ultimate_ga.php on line 524⏎GU

  4. #4 peak

    classroom may be a flipped classroom. In case you need a good example here is one from Edutopia.http://www.edutopia.org/blog/flip-stem-classroom-ainissa-ramirez and all flipped classrooms do

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. L20:344 strongly excites hyphen inside compound words
  2. L20:1886 strongly suppresses
  3. L19:804 excites
  4. L16:19 suppresses
  5. L23:474 excites
  6. L20:961 suppresses
  7. L19:171 suppresses
  8. L1:217 suppresses
  9. L4:490 suppresses decade digit in years like 1930s
  10. L22:2384 suppresses
  11. L0:1267 suppresses
  12. L16:471 suppresses blank spot before a year
  13. L19:1016 excites
  14. L25:422 excites
  15. L16:0 suppresses
  16. L24:248 excites
  17. L3:101 suppresses
  18. L24:318 suppresses
  19. L4:46 suppresses
  20. L2:964 suppresses
  21. L19:89 excites
  22. L24:1223 suppresses
  23. L2:422 excites
  24. L19:216 excites

Feeds later layers

  1. L27:1749 strongly excites hyphen joining word parts
  2. L27:261 strongly suppresses
  3. L27:183 suppresses
  4. L27:458 suppresses
  5. L27:2439 suppresses slash inside web addresses
  6. L27:184 suppresses
  7. L27:657 weakly suppresses
  8. L27:334 weakly suppresses
  9. L27:2657 weakly suppresses
  10. L27:1450 weakly suppresses
  11. L27:2224 weakly suppresses
  12. L27:1196 weakly suppresses
  13. L27:139 weakly suppresses
  14. L27:782 weakly excites
  15. L27:981 weakly suppresses
  16. L27:817 weakly suppresses
  17. L27:327 weakly suppresses
  18. L27:2373 weakly suppresses
  19. L27:283 weakly suppresses the spot right before a number
  20. L27:1147 weakly suppresses
  21. L27:627 weakly suppresses
  22. L27:740 weakly suppresses
  23. L27:913 weakly suppresses
  24. L27:858 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.

361 values between -3.69 and -0.153527 values between -0.15 and 3.406 values between 3.40 and 6.952 values between 6.95 and 10.504 values between 10.50 and 14.052 values between 14.05 and 17.602 values between 21.15 and 24.702 values between 24.70 and 28.241 values between 113.41 and 116.961 values between 134.70 and 138.25offnegativepositiveoften

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. :
  2. NULL
  3. odd
  4. ��
  5. .Menu
  6. occ
  7. ulan
  8. 調查
  9. ombs
  10. �

Suppresses

  1. diff
  2. dam
  3. me
  4. sm
  5. cr
  6. smart
  7. writing
  8. dom
  9. done
  10. re