Neuron Explorer

Qwen3-0.6B

words inside dialogue and questions

A neuron in layer 15 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 inside spoken lines and questions, especially on words like 'you' in 'Do you' or on the quote mark after a speaker's words. Turning it up made the model drift into unrelated topics, so the link is not clear.

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

    garden.⏎‘What are you doing here?’ he cried in a very gruff voice, and the children ran away. ‘

  2. #2 peak

    ship of the desert - cannot be far behind either. Do you want to learn more? Stay with us at AnimalWised and discover the native

  3. #3 peak

    immediately. Just as I say, “ Imagine the tip of your left pinky finger… where the nail meets the skin…” Do you think of it? Did you think of your pinky because I mentioned it?

  4. #4 peak

    drawing a picture of.⏎"God," he said.⏎"Why Johnny," she said, "you can't do thatbecause nobody knows what God looks like."⏎"They will when

  5. #5 peak

    to sit down with them, explain what you are trying to accomplish, and get them to start preparing for the next day is after

  6. #6 peak

    beloved local garden guru Carolyn Singer wrote, “’Grandma, why don’t you grow yucky flowers?’ ‘Why would I do that, Marcus?’ I asked, respectful of his three-year-old wisdom. ‘Well,’ he replied, ‘

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 need to get going and I need to get back to work and I should just give myself a go and do my job and I should have a good go at it and move on and I should have a go at it and just have

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

    ⏎Open Browser and type “”http://Server-IP/drupal”” which will open below screen to configure Drupal. Select ‘Standard‘, click on Save and Continue.⏎Choose language, click on Save and Continue⏎Fill up the details

  2. #2 peak

    were done to ensure that the dogs would not pull a lever just for fun or for fear of the unfamiliar dog. They also were shown how to give themselves a treat.⏎“This control excludes the possibility that the dogs did not pull on the tray out

  3. #3 peak

    lot of people, Ginger.⏎I would say I think … yeah. I think you’re talking about more than 80 million people.⏎Wow. Yeah. Okay. Can you define it for us? Can you describe what insulin resistance is in

  4. #4 peak

    this. All this time, they’ve been moping about in existential despair. What, did you think I didn’t write this blog to amuse myself? Okay, obviously, the paper is about how black flies might be of some use to

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. L11:2261 suppresses
  2. L10:375 excites
  3. L10:1003 weakly excites
  4. L1:2575 weakly suppresses
  5. L12:2835 weakly excites
  6. L14:2753 weakly suppresses
  7. L10:615 weakly excites
  8. L7:1893 weakly excites
  9. L6:1847 weakly suppresses
  10. L9:2336 weakly excites
  11. L13:2922 weakly excites
  12. L11:1030 weakly suppresses
  13. L12:2767 weakly excites
  14. L10:706 weakly excites
  15. L13:1259 weakly excites
  16. L11:561 weakly excites
  17. L2:2491 weakly suppresses
  18. L8:1071 weakly excites
  19. L13:45 weakly excites
  20. L2:748 weakly excites
  21. L12:530 weakly suppresses
  22. L12:2844 weakly excites
  23. L1:1648 weakly suppresses
  24. L9:1747 weakly suppresses

Feeds later layers

  1. L16:2828 suppresses
  2. L18:1260 suppresses
  3. L16:886 excites
  4. L23:1010 weakly suppresses
  5. L20:3056 weakly excites
  6. L19:2603 weakly suppresses
  7. L17:1528 weakly suppresses
  8. L22:2470 weakly suppresses
  9. L19:1973 weakly suppresses
  10. L16:2197 weakly excites
  11. L17:780 weakly suppresses
  12. L16:2377 weakly excites glosses, translations and unit conversions
  13. L18:2076 weakly suppresses
  14. L20:1055 weakly suppresses
  15. L17:2151 weakly suppresses
  16. L20:2972 weakly suppresses
  17. L24:2891 weakly excites
  18. L19:2425 weakly excites
  19. L16:3043 weakly suppresses
  20. L21:965 weakly excites
  21. L18:1224 weakly suppresses
  22. L20:2466 weakly excites
  23. L16:1247 weakly suppresses
  24. L16:2664 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 -0.69 and -0.643 values between -0.64 and -0.601 values between -0.60 and -0.563 values between -0.56 and -0.5211 values between -0.52 and -0.4811 values between -0.48 and -0.4418 values between -0.44 and -0.3951 values between -0.39 and -0.3555 values between -0.35 and -0.31107 values between -0.31 and -0.27154 values between -0.27 and -0.23221 values between -0.23 and -0.19248 values between -0.19 and -0.14348 values between -0.14 and -0.10414 values between -0.10 and -0.06514 values between -0.06 and -0.02590 values between -0.02 and 0.02380 values between 0.02 and 0.06266 values between 0.06 and 0.11182 values between 0.11 and 0.15120 values between 0.15 and 0.1982 values between 0.19 and 0.2365 values between 0.23 and 0.2723 values between 0.27 and 0.3115 values between 0.31 and 0.3613 values between 0.36 and 0.407 values between 0.40 and 0.443 values between 0.44 and 0.481 values between 0.61 and 0.651 values between 0.94 and 0.98offnegativepositiveoften

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. oser
  2. 特殊
  3. 独特
  4. السن
  5. Mana
  6. Savage
  7. _male
  8. 一方
  9. avery
  10. 特殊的

Suppresses

  1. sucesso
  2. .restart
  3. Congratulations
  4. thanked
  5. xes
  6. retries
  7. .Quit
  8. 啶
  9. lingen
  10. 通知