Neuron Explorer

Qwen3-0.6B

the 'way' in 'all the way'

A neuron in layer 2 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 responds to the word "way" in the phrase "all the way", as in "go all the way" or "all the way back to". Turning it up did not produce any obvious change in the text the model wrote.

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

    .⏎Go to Brightness/Contrast and set the contrast up all the way. Adjust brightness however you see fit. The idea is that you want the image

  2. #2 peak

    perform at a higher level.⏎All teachers at Washington Middle School are trained in the disciplinary literacy approach, starting at level one, and the master teachers go all the way to level three. “The number-one result we see is that kids<|endoftext|>About

  3. #3 peak

    a multiplicity of routes such that the system could survive a nuclear attack. It was not to keep the system’s data safe. And as ARPANET gave way to the purely academic world of the early internet, the idealistic graduate students who

  4. #4 peak

    riskiest foods.⏎While government and industry are helping to do their part to protect Americans, food safety can’t end with the FDA. It must continue all the way to our dinner tables, which means that we, as consumers, need to know

  5. #5 peak

    . Even if you don't make it all the way to the top of the mountain or hike all 40 miles of the trail

  6. #6 peak

    in higher education, making education more relevant in the dynamic global knowledge economy,<|endoftext|>Now, you'll notice that the edges of your picture have dashes going all the way up to the corners. Yes, unfortunately when the selected ellipse was turned into a

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: I have been told by someone: and I have been told by someone. What could that be? (I think of something you have been told by someone) is a common phrase. The word could be a common phrase. The word is

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

    Moving||5 Ways to Get Your Dog with Mobility Issues Moving||5 Exercise Tips for Arthritic Cats|⏎|Top Ten 'Small Breed' Dogs||Top Ten Dogs for Kids||If Your Pet Had Thumbs...||Top Five Pet Po

  2. #2 peak

    long as we do not lock our brains into the routine. Here are 10 tips to keep your brain active during all the years of your life.⏎Top 10 Tips to Keeping Your Brain Active.⏎1- Change Your Habits⏎

  3. #3 peak

    .⏎The Society was founded in 1956.<|endoftext|>- THE MAGAZINE⏎- NEW PRODUCTS⏎- CI Advanced Microsite⏎- CI Top 10⏎- Raw & Manufactured Materials Overview⏎- Classifieds & Services

  4. #4 peak

    Kingdom, with its constituent parts, including England, Scotland, Wales and Northern Ireland is an island country in northwest Europe. England, one of the world’s top 20 most developed economies, is home to The Beatles, the birthplace of

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. L0:1677 weakly excites
  2. L0:1119 weakly suppresses
  3. L0:1494 weakly suppresses
  4. L1:487 weakly suppresses
  5. L1:2988 weakly excites
  6. L1:1376 weakly suppresses
  7. L1:2215 weakly excites word completing an institution name
  8. L1:2898 weakly excites
  9. L1:626 weakly suppresses
  10. L0:988 weakly excites
  11. L0:1815 weakly suppresses
  12. L1:2369 weakly suppresses
  13. L1:189 weakly suppresses
  14. L1:619 weakly suppresses
  15. L1:788 weakly suppresses
  16. L0:1219 weakly excites
  17. L1:909 weakly excites
  18. L1:1244 weakly suppresses "of" after "board" or "off"
  19. L0:92 weakly suppresses
  20. L0:1691 weakly suppresses
  21. L1:1485 weakly excites
  22. L0:2454 weakly excites
  23. L1:2102 weakly suppresses
  24. L0:438 weakly suppresses

Feeds later layers

  1. L7:1463 weakly suppresses
  2. L3:981 weakly suppresses
  3. L15:272 weakly suppresses
  4. L20:1202 weakly suppresses
  5. L17:2542 weakly suppresses
  6. L17:1653 weakly excites
  7. L18:1980 weakly suppresses
  8. L5:19 weakly excites
  9. L16:1823 weakly excites
  10. L5:2150 weakly excites
  11. L8:414 weakly excites
  12. L10:808 weakly suppresses
  13. L6:1903 weakly excites
  14. L3:366 weakly excites
  15. L12:1241 weakly suppresses
  16. L14:1479 weakly suppresses
  17. L5:1997 weakly excites
  18. L5:2216 weakly suppresses
  19. L12:2364 weakly excites
  20. L4:2503 weakly excites
  21. L4:2553 weakly excites
  22. L8:2493 weakly suppresses
  23. L6:734 weakly suppresses
  24. L21:1218 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 -1.50 and -1.441 values between -1.44 and -1.371 values between -1.31 and -1.251 values between -0.87 and -0.811 values between -0.81 and -0.741 values between -0.74 and -0.681 values between -0.68 and -0.622 values between -0.62 and -0.554 values between -0.55 and -0.491 values between -0.49 and -0.431 values between -0.43 and -0.3718 values between -0.37 and -0.3031 values between -0.30 and -0.24114 values between -0.24 and -0.18254 values between -0.18 and -0.11595 values between -0.11 and -0.051160 values between -0.05 and 0.01864 values between 0.01 and 0.08497 values between 0.08 and 0.14213 values between 0.14 and 0.2090 values between 0.20 and 0.2723 values between 0.27 and 0.3320 values between 0.33 and 0.393 values between 0.39 and 0.453 values between 0.45 and 0.521 values between 0.52 and 0.582 values between 0.58 and 0.641 values between 0.64 and 0.712 values between 0.90 and 0.962 values between 0.96 and 1.02offnegativepositiveoften

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. alytics
  2. 级以上
  3. Vimeo
  4. Prop
  5. clutter
  6. 申报
  7. frau
  8. 瑕
  9. {:
  10. sterdam

Suppresses

  1. ream
  2. piel
  3. REAM
  4. 对他们
  5. омн
  6. İN
  7. スタン
  8. рег
  9. منت
  10. نك