Neuron Explorer

Qwen3-0.6B

commas after a 'since' clause

A neuron in layer 13 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 fires on the comma that ends a 'since...' or similar reason clause, where the main point is about to follow. Turning it up made the model drift toward talking about reasons and ideas, while turning it down gave less clear results.

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

    even at the expense of personal well-being. It can, however, be difficult to separate MIT’s culture from its people. Since the people help create the culture, one cannot ignore the fact that MIT’s students tend to be internally driven and place

  2. #2 peak

    the Fed injects more dollars into the economy it will show up as increased final demand/consumption and boost GDP. Since printing more money doesn’t create wealth, that doesn’t tell us much about our economic health (e.g., spending bo

  3. #3 peak

    what comprises their skeletons. Calcarea sponges have limestone skeletons. Hexactinellida have skeletons made of silica. (Since glass is made from silica, these sponges are nicknamed "glass sponges." ) Scleros

  4. #4 peak

    ’s new queen likely settled on the jet’s nozzle to rest (hives often migrate to new locations with new queens) and since the hive never leaves its queen, everyone decided to have a kick-back right on the nose of the military’s sh

  5. #5 peak

    automation, robotics and many more. You can simply browse in the menu and on the page or leave a comment. Since new Raspberry Pi instructions and tutorials appear regularly, it is definitely worth activating the newsletter.⏎WiringPi is a library that is

  6. #6 peak

    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 no need to unless you want everything to look consistent<|endoftext|>Whether it

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: The first of the three: The first is a number, and it is a number in the space, and it is a number: The first is a number: The first is a number: The first is a number: The first 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

    to the crest of the ridge where the best huckleberries grew.⏎Some years we got lucky. Every bush would be loaded, and we could be choosy—sampling deep

  2. #2 peak

    The studio probably had little choice but to create a form letter.⏎I don’t think the form letter nature of the rejection is necessarily impersonal. It sticks to objective criticisms. Probably you can’t pin a rejection on a

  3. #3 peak

    overwintering, and diapause—an insect version of hibernation.”⏎For example, “Boll weevils were thought to feed on pollen only, but my work reveals otherwise,” he says.⏎Nectars Could Be Where It

  4. #4 peak

    too dangerous to fly. Her flight home to Vancouver had been scratched because it connected through Frankfurt, an aviation super-hub shuttered because of the overhanging haze. Hickson, who works for a private geothermal energy company, acknowledged 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. L12:815 excites
  2. L12:801 excites
  3. L8:191 suppresses
  4. L11:2171 suppresses
  5. L12:731 excites
  6. L10:1677 suppresses
  7. L7:84 excites
  8. L12:834 excites
  9. L12:935 excites
  10. L11:2719 suppresses
  11. L10:2981 suppresses
  12. L8:309 suppresses
  13. L10:2621 excites
  14. L12:1672 excites
  15. L10:766 suppresses
  16. L11:2305 weakly excites
  17. L10:422 weakly suppresses
  18. L12:2718 weakly suppresses
  19. L4:2526 weakly excites
  20. L12:1796 weakly suppresses
  21. L7:766 weakly excites
  22. L11:2733 weakly suppresses
  23. L2:2373 weakly excites
  24. L5:246 weakly suppresses

Feeds later layers

  1. L14:1740 excites
  2. L14:1433 excites
  3. L14:756 suppresses
  4. L18:2592 suppresses comma after an opening clause
  5. L16:1393 suppresses
  6. L14:168 suppresses
  7. L16:2528 suppresses
  8. L14:848 excites
  9. L16:1702 weakly suppresses
  10. L15:997 weakly excites
  11. L15:2551 weakly excites
  12. L24:1429 weakly excites
  13. L15:1189 weakly suppresses
  14. L14:2212 weakly suppresses
  15. L16:2386 weakly suppresses
  16. L15:1847 weakly excites
  17. L14:1704 weakly excites
  18. L24:612 weakly suppresses
  19. L14:749 weakly excites
  20. L21:1490 weakly excites
  21. L15:318 weakly suppresses
  22. L16:1838 weakly suppresses
  23. L15:2328 weakly excites
  24. L16:861 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.

2 values between -2.41 and -2.274 values between -2.13 and -1.981 values between -1.70 and -1.562 values between -1.56 and -1.425 values between -1.42 and -1.287 values between -1.28 and -1.1315 values between -1.13 and -0.9916 values between -0.99 and -0.8526 values between -0.85 and -0.7143 values between -0.71 and -0.5780 values between -0.57 and -0.43170 values between -0.43 and -0.29379 values between -0.29 and -0.141346 values between -0.14 and -0.001362 values between -0.00 and 0.14220 values between 0.14 and 0.2892 values between 0.28 and 0.4267 values between 0.42 and 0.5631 values between 0.56 and 0.719 values between 0.71 and 0.8510 values between 0.85 and 0.995 values between 0.99 and 1.133 values between 1.13 and 1.275 values between 1.27 and 1.413 values between 1.41 and 1.561 values between 1.70 and 1.843 values between 1.84 and 1.981 values between 3.11 and 3.25offnegativepositiveoften

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. ault
  2. amenti
  3. evice
  4. cq
  5. istine
  6. -ret
  7. ody
  8. ewed
  9. ave
  10. lica

Suppresses

  1. |
  2. 部分
  3. IN
  4. 的部分
  5. .parts
  6. day
  7. part
  8. |^
  9. parts
  10. выраж