Neuron Explorer

Qwen3-0.6B

'sure' in 'are you sure' phrasing

A neuron in layer 9 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 strongly on the word 'sure' when it follows phrases like 'if you are not' or 'once you're'. It also responds to 'know' in 'how do you know'. Turning it up did not clearly change the topic of the writing.

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

    We have spent thousands of hours working with students and developing strategies to help them to succeed on the test. Our track record demonstrates just that. If you are not sure if the

  2. #2 peak

    How to deal with it⏎The first step is to identify the problem. When you are sure that your child has a bully problem, you will need to alarm the teacher about

  3. #3 peak

    is an ignorant man. When he tells you he has a hunch about how it is going to work, he is uncertain about it. When he is pretty sure of how it is going to work, and he tells you, "This is

  4. #4 peak

    possible. I don't think there is another equation that relates the variables in the first one. If there was you could solve them simultaneously.⏎What makes you so sure that the different forms of DNA have different extintion coefficients? For calculating the

  5. #5 peak

    , you'll need to disconnect from it immediately. That's to ensure that your infected PC does not pass the bug on to any other computer. Once you're sure your virus protection program has the latest updates, disconnect from the Internet as well.

  6. #6 peak

    . And that’s part of the problem: it’s so obvious that it seems like it probably wouldn’t work at all. For one thing, how do you know if your

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: a set of 2000 items, a set of 200 items, and a set of 2000 items, a set of 200 items, and a set of 2000

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

    County, California Business Council, show 3 to 5% of children in Orange County have ADHD (Attention Deficit Hyper Activity). The U.S. Census Bureau published a report in 2007 stating nearly 10,0

  2. #2 peak

    dating uses radioactive decay rates for radiometric dating.Start studying relative dating. Explain the radiometric dating, this method involves placing ge

  3. #3 peak

    and find single man younger woman. Determining when the fossils and contrast between relative abundances of fossil dating could be applied to answer the field guide. Radiometric dating methods is true about index fossils in the fossils read more how. Scientific measurements such

  4. #4 peak

    may account for anywhere from 5% to 10% of autism in families with only one autistic child, the researchers said.⏎“These genetic studies open the door to understanding the biology and pathophysiology of the disorder,” says Dr. Matthew

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. L8:89 suppresses
  2. L6:2618 excites
  3. L8:3030 excites
  4. L8:1164 weakly suppresses
  5. L8:1589 weakly excites
  6. L6:2132 weakly suppresses
  7. L8:60 weakly excites
  8. L8:113 weakly suppresses
  9. L7:1883 weakly suppresses
  10. L8:1926 weakly suppresses
  11. L6:611 weakly suppresses
  12. L7:2789 weakly suppresses
  13. L8:671 weakly excites
  14. L7:2416 weakly suppresses
  15. L4:2649 weakly suppresses
  16. L8:1492 weakly suppresses
  17. L4:929 weakly excites
  18. L6:122 weakly suppresses
  19. L7:1368 weakly excites
  20. L7:2696 weakly suppresses
  21. L8:3008 weakly suppresses
  22. L6:229 weakly excites
  23. L6:2903 weakly suppresses
  24. L8:161 weakly suppresses

Feeds later layers

  1. L10:1236 excites
  2. L16:2483 excites
  3. L11:1774 excites
  4. L12:2517 excites verbs of finding out or thinking
  5. L18:2070 excites
  6. L20:533 weakly excites
  7. L24:1977 weakly excites
  8. L12:1947 weakly excites
  9. L18:1993 weakly excites
  10. L12:1816 weakly excites
  11. L17:307 weakly excites
  12. L20:1215 weakly suppresses
  13. L11:54 weakly excites
  14. L10:1011 weakly excites
  15. L10:1591 weakly excites
  16. L12:292 weakly excites
  17. L10:562 weakly excites
  18. L12:1223 weakly excites
  19. L18:2122 weakly suppresses
  20. L21:481 weakly suppresses
  21. L22:1707 weakly suppresses
  22. L18:666 weakly suppresses
  23. L22:1823 weakly excites
  24. L11:2519 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 -1.67 and -1.531 values between -1.53 and -1.391 values between -1.39 and -1.261 values between -1.26 and -1.121 values between -1.12 and -0.995 values between -0.99 and -0.8517 values between -0.85 and -0.7214 values between -0.72 and -0.5827 values between -0.58 and -0.4567 values between -0.45 and -0.31240 values between -0.31 and -0.18865 values between -0.18 and -0.042118 values between -0.04 and 0.09398 values between 0.09 and 0.2361 values between 0.23 and 0.3727 values between 0.37 and 0.5017 values between 0.50 and 0.6413 values between 0.64 and 0.7710 values between 0.77 and 0.917 values between 0.91 and 1.045 values between 1.04 and 1.181 values between 1.18 and 1.311 values between 1.31 and 1.452 values between 1.45 and 1.581 values between 1.58 and 1.721 values between 1.72 and 1.851 values between 1.99 and 2.123 values between 2.12 and 2.261 values between 2.26 and 2.401 values between 3.61 and 3.75offnegativepositiveoften

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. 毕竟
  2. ABOUT
  3. ampion
  4. xứ
  5. 一二
  6. ѓ
  7. 危险
  8. abilit
  9. -about
  10. ABILITY

Suppresses

  1. Networks
  2. democracy
  3. networks
  4. odes
  5. LM
  6. ooled
  7. hoses
  8. hashtag
  9. دان
  10. ész