Neuron Explorer

Qwen3-0.6B

unclear: blank before a year or number

A neuron in layer 11 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 weakly on a blank spot that sits just before a year or a number, often in a citation or date. The signal is faint, so it is hard to say what it really does. Turning it up made the text drift into lists of numbers and years.

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

    introduced. As an added⏎bonus, an introduction to JShell, Java's interactive programming⏎tool, is included.⏎|Java: A Beginner's Guide, 9

  2. #2 peak

    aanews.noaa.gov⏎– Erich Luening⏎Print this page<|endoftext|>Today is Sunday, Feb. 3, the 34th day of 2013. There are 331 days left in the year

  3. #3 peak

    modern conflict, writes Audrey Borowski.⏎A Theory of the Drone. Grégoire Chamayou. Trans. by Janet Lloyd. The New Press. 2015.⏎In the past decade, the deployment of hunter-killer

  4. #4 peak

    es wrought by the present technological age.⏎Michael Freeman, Bathurst Heights Secondary School, North York, ON.⏎1971-1979 | 1980-1985 | 1986-

  5. #5 peak

    is not (Shectman & Horowitz, 2003). Studies find that people do not treat computers as people outside

  6. #6 peak

    , 19. Martin Beatty, 20. Patt Carr.⏎Sources: NAI, (National Archives of Ireland

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: " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " "

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

    's life.⏎Methane has been present in much larger quantities in past geological ages. And large quantities of it exist in combination on the ocean beds.⏎Its greenhouse gas effect is felt to be around twenty-five times that or carbon dioxide, and

  2. #2 peak

    泰国学生在学习汉语助动词中的困难及偏误情况,分析他们产生偏误的原因。我们发现经过基于对比研究设计的汉语助动词课堂教学后,“后测(Post-test)”的选择正确率高于

  3. #3 peak

    — that the Earth stood still in outer space while the sun and other celestial bodies rotated around it. Such findings outraged the Roman Catholic Church.⏎One problem with Copernicus’ observations

  4. #4 peak

    , and preferably 1.5 degrees Celsius, from pre-industrial levels. The aim is to use the best available scientific data to reduce the global amount of greenhouse gases, mainly carbon dioxide and methane emissions, in the Earth’s atmosphere.⏎Leading

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. L8:7 excites
  2. L8:16 suppresses
  3. L8:3 excites
  4. L2:36 excites
  5. L6:12 excites
  6. L8:10 excites
  7. L8:4 excites
  8. L6:7 suppresses
  9. L5:46 excites
  10. L3:59 excites
  11. L8:14 excites
  12. L5:503 excites
  13. L9:5 excites
  14. L5:117 excites
  15. L8:13 suppresses
  16. L2:70 excites
  17. L5:315 suppresses
  18. L7:66 excites
  19. L0:1301 weakly suppresses
  20. L4:19 weakly suppresses
  21. L7:21 weakly excites
  22. L1:1340 weakly suppresses
  23. L5:198 weakly excites unclear: weak "and" in lists
  24. L2:1710 weakly suppresses

Feeds later layers

  1. L12:130 excites
  2. L13:41 excites the word 'the' after a lead-in phrase
  3. L14:31 excites
  4. L12:4 excites document-end markers, mixed
  5. L14:167 excites
  6. L17:0 excites end-of-document token boundary
  7. L27:47 excites
  8. L27:81 excites
  9. L16:828 excites
  10. L27:386 excites
  11. L14:12 excites
  12. L14:934 excites
  13. L14:22 excites
  14. L27:39 excites
  15. L14:17 excites
  16. L14:10 excites
  17. L14:25 excites
  18. L27:55 excites
  19. L12:403 excites
  20. L14:14 excites
  21. L14:1 excites
  22. L27:314 excites
  23. L25:11 excites
  24. L14:21 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.

35 values between -0.01 and 0.053 values between 0.05 and 0.112 values between 0.11 and 0.178 values between 0.17 and 0.2316 values between 0.23 and 0.2923 values between 0.29 and 0.3552 values between 0.35 and 0.4164 values between 0.41 and 0.4695 values between 0.46 and 0.52129 values between 0.52 and 0.58142 values between 0.58 and 0.64181 values between 0.64 and 0.70227 values between 0.70 and 0.76274 values between 0.76 and 0.82248 values between 0.82 and 0.87309 values between 0.87 and 0.93277 values between 0.93 and 0.99227 values between 0.99 and 1.05232 values between 1.05 and 1.11244 values between 1.11 and 1.17210 values between 1.17 and 1.23176 values between 1.23 and 1.29175 values between 1.29 and 1.34137 values between 1.34 and 1.40103 values between 1.40 and 1.4675 values between 1.46 and 1.5267 values between 1.52 and 1.5858 values between 1.58 and 1.6449 values between 1.64 and 1.7022 values between 1.70 and 1.7514 values between 1.75 and 1.818 values between 1.81 and 1.8711 values between 1.87 and 1.934 values between 1.93 and 1.992 values between 1.99 and 2.056 values between 2.05 and 2.112 values between 2.11 and 2.161 values between 2.28 and 2.34offnegativepositiveoften

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. ies
  2. (
  3. f
  4. 1
  5. cap
  6. op
  7. ational
  8. ou
  9. 2
  10. 3

Suppresses

  1. ATUS
  2. Crimea
  3. 射手
  4. SHOT
  5. 自然是
  6. sembled
  7. flowed
  8. _Statics
  9. BOSE
  10. COURT