Neuron Explorer

Qwen3-0.6B

word or phrase being named or defined

A neuron in layer 16 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron tends to light up on the key word a sentence is about, such as 'focus', 'trees', or 'testosterone', especially where the text is pointing at a topic. Turning it up made the writing wander into unrelated subjects like 3D printers and chess.

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

    C F O S U.⏎No direct anagrams for focus found in this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with focus by next letter<|endoftext|>I was

  2. #2 peak

    workings of decision tree algorithms, explaining how they use data to make predictions and offering a comprehensive understanding of this fascinating topic. Get ready to explore the world of decision trees and discover how they make predictions that are accurate, reliable, and effective.⏎II

  3. #3 peak

    To Fiber Optics Technology Understanding how fiber optics are made and function for uses in everyday life is an intriguing work of art combined with science. Fiber optics

  4. #4 peak

    system as a whole to appear sluggish to the user. It is difficult to have a goal on overall sluggishness of the system, so having a goal on memory consumption is reasonable.⏎When defining your performance goals, take into consideration the perceived size of

  5. #5 peak

    30 percent of men who sought treatment for male infertility were discovered to have low testosterone levels.⏎Understanding what testosterone does for a man makes recognizing the signs of low testosterone easier. Lets break it down. First,

  6. #6 peak

    , caffeine, and artificial sweeteners used to jack up drinkers of diet sodas, and a new study indicating that drinking diet soft drinks may increase risk for stroke only makes us more worried about the mysterious effects of those calorie-free cans. Researchers tracked

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, what, and, what, and, what, and, what, and, what, and, what, and, what, and, what, and, what, and, what, and, what, and,

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

    major river in Flood Zone 3 is Calleguas Creek with a watershed area of approximately 341 square miles. All stream flows in Zone 3 eventually end up in Mugu Lagoon before entering the Pacific Ocean. Major trib

  2. #2 peak

    is among the oldest known crops dating back 10,000 years to Mexico and America. The name squash, which means uncooked or eaten raw, comes from the Native American word askutasquash.⏎Butternut squash tastes

  3. #3 peak

    meant by "hashing function"? How is it used in hash data clusters and hash joins?⏎By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.⏎A hashing function is

  4. #4 peak

    or spongy bone, light, porous bone enclosing numerous large spaces that give a honeycombed or spongy appearance. The bone matrix, or framework, is organized into a three-dimensional latticework of bony processes, called

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. L15:851 suppresses
  2. L15:404 suppresses
  3. L15:668 suppresses
  4. L14:1133 suppresses unclear: commas and nouns before an aside
  5. L15:1344 excites
  6. L15:161 suppresses
  7. L15:879 suppresses
  8. L13:2609 suppresses
  9. L15:1543 suppresses
  10. L15:435 suppresses
  11. L15:2831 suppresses
  12. L15:827 suppresses
  13. L15:396 excites
  14. L13:706 excites
  15. L14:1060 excites
  16. L14:220 excites
  17. L15:870 excites
  18. L14:2651 suppresses
  19. L15:1255 excites
  20. L15:806 suppresses
  21. L15:1513 suppresses
  22. L14:1082 excites
  23. L15:386 excites
  24. L14:674 suppresses

Feeds later layers

  1. L20:330 suppresses
  2. L17:1028 excites
  3. L17:530 excites
  4. L18:2602 suppresses
  5. L19:1175 excites
  6. L18:1972 excites
  7. L19:363 excites unclear: table pipes and split word fragments
  8. L19:249 excites
  9. L19:1297 suppresses
  10. L19:340 suppresses
  11. L17:59 excites
  12. L17:2781 excites
  13. L19:1288 suppresses
  14. L20:25 excites
  15. L18:1154 excites
  16. L17:1065 excites
  17. L20:321 suppresses
  18. L19:146 excites
  19. L18:938 suppresses
  20. L20:914 excites
  21. L19:2074 suppresses
  22. L18:1128 suppresses
  23. L20:40 weakly excites
  24. L17:1171 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 -5.26 and -5.032 values between -3.18 and -2.953 values between -2.95 and -2.721 values between -2.49 and -2.261 values between -2.26 and -2.036 values between -2.03 and -1.7915 values between -1.79 and -1.569 values between -1.56 and -1.3321 values between -1.33 and -1.1031 values between -1.10 and -0.8779 values between -0.87 and -0.64296 values between -0.64 and -0.41750 values between -0.41 and -0.181821 values between -0.18 and 0.05560 values between 0.05 and 0.28147 values between 0.28 and 0.5273 values between 0.52 and 0.7538 values between 0.75 and 0.9820 values between 0.98 and 1.2112 values between 1.21 and 1.446 values between 1.44 and 1.678 values between 1.67 and 1.902 values between 1.90 and 2.132 values between 2.13 and 2.362 values between 2.36 and 2.591 values between 3.06 and 3.291 values between 3.75 and 3.98offnegativepositiveoften

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. нар
  2. implicitly
  3. Animator
  4. IMPLIED
  5. linkplain
  6. 蛲
  7. ignon
  8. XObject
  9. вин
  10. weg

Suppresses

  1. 但是他
  2. �
  3. )
  4. 때문이다
  5. 轰
  6. )/
  7. ieu
  8. çi
  9. _)
  10. 羁