Neuron Explorer

Qwen3-0.6B

'yes, you' style emphatic replies

A neuron in layer 7 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word after an emphatic 'Yes,' or 'No,' or 'That's right', mostly 'you' or 'that', as in 'Yep, you can still be fired'. Turning it up made the output wander into unrelated generic prose, so the link to the trigger is weak.

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

    citizens. There is no federal law outlawing discrimination nationwide leaving residents of some states unprotected against discrimination in employment, housing, and private or public services. Yep, you can still be fired, denied housing or refused service for

  2. #2 peak

    Yes, you heard it right!⏎How Low-Carb Diet Risks Heart-Health:⏎A study

  3. #3 peak

    strips and then crimp the bottom⏎- plant pots – if you do buy some transplants, keep the pots they come in and reuse them⏎Yes, you can also buy containers. But with all these free versions above, there really is

  4. #4 peak

    perfume.⏎On the other side of the world, the Mayans are believed to have utilized the gum from the manilkara chicle evergreen tree (yes, that’s the inspiration for “Chiclet”) in a

  5. #5 peak

    in its practices.⏎You may also like..⏎- Education and Training⏎- Special Populations⏎- Youth in Transition<|endoftext|>White Turkey – White Turkey⏎No, you’re not dreaming – White turkey are a real thing! White turkeys are bred

  6. #6 peak

    What did the payphone do to deserve this?”⏎“Well, the payphone didn’t do anything; the cell phone came along,” Thomas replied.⏎That’s right: our phones got so smart they began putting pay phones out of business.⏎“

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: "Add" and its arguments. I need to know the arguments for this. The user is trying to build a project where each "Add" is a "Add" with arguments. The user might have to add dependencies for the project.

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

    project by select File >> New >> Project. Select General >> Empty Project and type in the project name.⏎- Under folder Source, right-click Select Add >> New Item.⏎- Under Code, Select C++ File (.cpp), however, when you

  2. #2 peak

    until you have build solution.⏎- To build solution select Build >> Build Solution (F7)⏎- Running

  3. #3 peak

    ’t have cabal-install, you can manually download and install bool-extras:⏎tar xzf bool-extras-0.3.0.tgz⏎./Setup.lhs<|endoftext|>Education Pioneers recruits, develops, and connects professionals from diverse

  4. #4 peak

    the drive and folder into which you want to install the free C++Builder 5 command line tool development system.⏎- From the bin directory of your installation: Add “c:\Borland\Bcc55\bin” to the

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. L3:1234 suppresses
  2. L2:70 excites
  3. L3:59 excites
  4. L6:2304 weakly suppresses
  5. L5:1725 weakly suppresses
  6. L2:41 weakly suppresses
  7. L4:2152 weakly excites
  8. L3:2681 weakly suppresses
  9. L3:2704 weakly suppresses
  10. L6:1541 weakly excites
  11. L6:2152 weakly excites
  12. L6:292 weakly excites
  13. L6:2244 weakly suppresses
  14. L2:75 weakly suppresses
  15. L4:127 weakly suppresses
  16. L6:2074 weakly suppresses
  17. L2:1694 weakly excites
  18. L4:2177 weakly suppresses
  19. L3:1901 weakly suppresses
  20. L1:1430 weakly excites
  21. L2:1982 weakly suppresses
  22. L6:1923 weakly suppresses
  23. L4:2831 weakly suppresses
  24. L6:13 weakly excites

Feeds later layers

  1. L8:2403 suppresses
  2. L20:1363 suppresses
  3. L10:2564 weakly suppresses
  4. L23:1101 weakly excites
  5. L10:1314 weakly suppresses
  6. L14:1051 weakly suppresses
  7. L15:2841 weakly excites
  8. L8:665 weakly suppresses
  9. L23:1108 weakly suppresses
  10. L9:2588 weakly suppresses
  11. L9:2811 weakly suppresses
  12. L8:2657 weakly excites
  13. L8:2532 weakly excites
  14. L8:2172 weakly suppresses
  15. L8:2173 weakly excites
  16. L20:1953 weakly suppresses
  17. L9:2243 weakly suppresses
  18. L8:2540 weakly excites
  19. L17:2720 weakly excites
  20. L10:639 weakly suppresses
  21. L9:1775 weakly excites
  22. L21:686 weakly excites
  23. L8:1976 weakly excites
  24. L15:2611 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 -0.68 and -0.614 values between -0.61 and -0.547 values between -0.54 and -0.489 values between -0.48 and -0.4123 values between -0.41 and -0.3467 values between -0.34 and -0.27133 values between -0.27 and -0.21301 values between -0.21 and -0.14580 values between -0.14 and -0.07882 values between -0.07 and -0.00975 values between -0.00 and 0.07515 values between 0.07 and 0.13237 values between 0.13 and 0.20111 values between 0.20 and 0.2736 values between 0.27 and 0.3414 values between 0.34 and 0.407 values between 0.40 and 0.472 values between 0.47 and 0.541 values between 0.54 and 0.611 values between 0.81 and 0.881 values between 1.35 and 1.421 values between 1.96 and 2.03offnegativepositiveoften

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. uetype
  2. ogram
  3. 纠错
  4. êm
  5. activated
  6. ceeded
  7. ��
  8. umbed
  9. adamente
  10. esterday

Suppresses

  1. 工业大学
  2. (Seq
  3. Include
  4. 浡
  5. ting
  6. Creative
  7. ends
  8. #-}⏎⏎
  9. Include
  10. smith