Neuron Explorer

Qwen3-0.6B

opening parenthesis after a noun

A neuron in layer 10 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 an opening bracket that comes right after a noun, especially when the bracket is about to give a short alternative name, a size, or an abbreviation, such as '1 cup (8 ozs.)' or 'the Commission (the Commission)'. Turning it up did not make the text clearly about brackets; the output drifted into unrelated sentences.

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

    oxide in warm distilled water once to twice a week to remove even chronic yeast infections.⏎Colonic or Enema:⏎For a colonic, add 1 cup (8 ozs.) 3% H202 to 5 gallons warm

  2. #2 peak

    -sense considerations will make your trail-travel safer and more enjoyable for everyone:⏎- Share the trail. Show consideration for others.⏎- Please keep to the right except when passing⏎- Control your speed⏎- Warn others when passing⏎- Keep

  3. #3 peak

    soapy water so they are very clean. You can also disinfect them with a mild bleach solution.⏎- yoghurt containers – the small 8 ounce (250ml) single serving sized ones are

  4. #4 peak

    of protein function have undergone a marked transformation with the discovery of an aspect of enzyme evolution called "moonlighting." That is, a protein can acquire a second (moonlighting) function without concomitantly losing all or part of its

  5. #5 peak

    remedy given by the trial judge was improper. The process will now be remitted back to the planning stage that was previous to the Peel Watershed Planning Commission’s (the “Commission”) Final Recommended Report (“Final Recommended Plan”), instead of being rem

  6. #6 peak

    on the calyx and stem end of the fruit.<|endoftext|>The American Lung Association's (ALA's) Not-On-Tobacco (N-O-T) web site asserts

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 past, present, and future, the three of them, the three of them, the three, the three of them, the three, the three, the three, the three, the three, the three, the three, the

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

    ⏎geologic time frame relating to amber formation,⏎based on Jean Langenheim's work with amber/botanical affinities reproduced in⏎Rice: Amber the Golden Gem of the Ages, the Polish Academy of Sciences:⏎Amber in Nature

  2. #2 peak

    make best use of the earnings they can get out of their job, yet lots of traditional publishers choose to publish their books with a huge publishing house.⏎Many authors that select to self-publish their books locate typical publishers and also agents to be a valuable

  3. #3 peak

    from conflict to cyclone relief.”⏎Senior emergency response specialist, United Nations Office for the Coordination of Humanitarian Affairs⏎Cyclones Idai and Kenneth that swept across southern Africa in 2019 are examples of how the climate

  4. #4 peak

    in an easy and fun manner. In this course I do just that and with one of the most useful IDEs coupled with the Java programming language.⏎Java, a wonderfully simple programming language that has encouraged many programmers to pursue careers involving Java over any

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. L7:365 excites
  2. L7:685 excites
  3. L8:577 excites
  4. L8:444 excites
  5. L9:1903 suppresses
  6. L6:155 excites
  7. L7:616 suppresses
  8. L4:211 suppresses
  9. L9:738 weakly suppresses
  10. L9:865 weakly suppresses
  11. L6:342 weakly suppresses
  12. L7:1854 weakly suppresses
  13. L6:996 weakly suppresses
  14. L2:105 weakly excites
  15. L5:976 weakly excites
  16. L3:1193 weakly suppresses
  17. L9:132 weakly suppresses
  18. L2:1588 weakly excites
  19. L2:1564 weakly suppresses
  20. L4:89 weakly suppresses
  21. L2:89 weakly suppresses
  22. L6:829 weakly excites
  23. L6:217 weakly excites
  24. L8:309 weakly suppresses

Feeds later layers

  1. L11:578 suppresses
  2. L14:728 suppresses
  3. L15:304 excites
  4. L15:2792 suppresses
  5. L11:1327 suppresses
  6. L15:1551 excites
  7. L17:2216 suppresses
  8. L11:413 excites
  9. L14:602 suppresses
  10. L15:1571 suppresses
  11. L11:948 suppresses
  12. L12:731 weakly suppresses
  13. L16:731 weakly excites
  14. L11:782 weakly excites
  15. L11:398 weakly excites
  16. L13:652 weakly excites
  17. L13:628 weakly suppresses
  18. L14:1487 weakly suppresses
  19. L15:681 weakly suppresses
  20. L12:2605 weakly suppresses
  21. L14:850 weakly suppresses
  22. L14:2402 weakly suppresses
  23. L15:2568 weakly suppresses
  24. L13:1914 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 -3.91 and -3.771 values between -3.62 and -3.481 values between -3.04 and -2.891 values between -2.89 and -2.751 values between -2.45 and -2.311 values between -2.16 and -2.023 values between -1.87 and -1.721 values between -1.72 and -1.581 values between -1.58 and -1.432 values between -1.43 and -1.285 values between -1.28 and -1.147 values between -1.14 and -0.997 values between -0.99 and -0.8516 values between -0.85 and -0.7014 values between -0.70 and -0.5542 values between -0.55 and -0.4185 values between -0.41 and -0.26378 values between -0.26 and -0.121798 values between -0.12 and 0.03998 values between 0.03 and 0.18359 values between 0.18 and 0.32118 values between 0.32 and 0.4735 values between 0.47 and 0.6216 values between 0.62 and 0.763 values between 0.76 and 0.914 values between 0.91 and 1.055 values between 1.05 and 1.202 values between 1.20 and 1.351 values between 1.35 and 1.491 values between 1.64 and 1.781 values between 1.78 and 1.93offnegativepositiveoften

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. 其中之一
  2. offs
  3. hesion
  4. hoped
  5. 每位
  6. iswa
  7. є
  8. 鼓励
  9. ixture
  10. minLength

Suppresses

  1. Presidency
  2. <header
  3. '''⏎⏎
  4. uw
  5. RemoteException
  6. zn
  7. ippet
  8. ',⏎
  9. 上调
  10. ',⏎⏎