Neuron Explorer

Qwen3-0.6B

plural '(s)' after a noun

A neuron in layer 7 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 lights up on the '(s)' that writers add after a noun to show it could be singular or plural, as in 'type(s)' or 'animal(s)'. Turning it up did not produce text that clearly reflects this idea.

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

    and delayed the release? Provide reasons for your decision.⏎You must first post your answer(s) before you can see the other students’ posts. Submit your initial post(s

  2. #2 peak

    American Society for the Prevention of Cruelty to Animals (ASPCA), "Animals in Entertainment" refers to any animal(s) used to act, perform, fight and/or kill for the enjoyment of humans

  3. #3 peak

    case of a power outage or other emergency. Furthermore, employers may want to designate a person(s) responsible for obtaining and maintaining the SDSs. If the employer does not have

  4. #4 peak

    include generalized anxiety disorder, social anxiety disorder, and separation anxiety disorder. It is possible to have more than one, or possibly all three. No matter what type(s) of anxiety you are experiencing.⏎Examples of anxiety disorders include generalized anxiety disorder,

  5. #5 peak

    , you may have dry eye. A thorough ocular surface evaluation with our dry eye specialists will determine whether dry eye is present, and if so, what type(s) of dry eye are affecting

  6. #6 peak

    permanent magnet or an electromagnet.⏎A typical docking operation begins with the docking spacecraft maneuvering into position near the docking port on the parent vehicle. The electromagnet(s) are then turned on, and, if necessary, the docking spacecraft is then

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, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and,

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

    wolves.⏎tAsAM ruditashabdena gavAM haMbhAraveNa cha |⏎vrajasyotthApanaM chakrurghoShavR^iddhAH

  2. #2 peak

    studies in more clinically relevant models. Proceedings of the National Academy of Sciences USA (PNAS) has published the study online this week.⏎Argentinian scientist Andrés Dekanty, a Juan de la Cierva researcher in the Mil

  3. #3 peak

    igulinœ and genus Clangula; a garrot. The common goldeneye is C. glaucion or C. clangula of Europe and Americn. Barrow's goldeneye is the Rocky Mountain garrot, C. bar

  4. #4 peak

    ordinary Muslims gathered in every street, whipped into a frenzy by slogans which rang out from the loudspeakers of mosques. One such slogan was,⏎‘Asi gachchi Pakistan, Batao roas te Batanev san

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. L3:1291 suppresses
  2. L3:720 excites
  3. L5:1577 excites
  4. L2:1599 excites
  5. L6:949 excites
  6. L5:2843 excites
  7. L3:1016 excites
  8. L3:494 suppresses
  9. L4:1044 excites
  10. L2:1301 suppresses
  11. L2:2129 weakly suppresses
  12. L4:2807 weakly excites
  13. L3:284 weakly suppresses
  14. L3:1092 weakly suppresses
  15. L6:2181 weakly suppresses
  16. L3:817 weakly suppresses
  17. L2:1397 weakly excites
  18. L3:1955 weakly excites
  19. L4:1088 weakly suppresses
  20. L4:290 weakly excites
  21. L0:2566 weakly suppresses
  22. L3:27 weakly excites
  23. L4:1067 weakly excites
  24. L4:1300 weakly suppresses

Feeds later layers

  1. L9:897 suppresses
  2. L8:517 suppresses
  3. L10:799 suppresses
  4. L11:1722 suppresses
  5. L15:998 suppresses period inside an abbreviation
  6. L8:917 suppresses
  7. L15:1498 suppresses
  8. L9:39 weakly suppresses
  9. L14:143 weakly excites 'the' inside journal and institution names
  10. L11:2942 weakly suppresses
  11. L9:2354 weakly suppresses
  12. L11:1393 weakly suppresses
  13. L11:705 weakly excites
  14. L10:77 weakly suppresses
  15. L12:753 weakly excites
  16. L15:1382 weakly suppresses
  17. L9:2249 weakly suppresses
  18. L15:1554 weakly suppresses
  19. L13:1210 weakly suppresses
  20. L15:991 weakly suppresses
  21. L11:765 weakly suppresses
  22. L9:392 weakly suppresses "and" inside organisation names
  23. L17:1612 weakly suppresses
  24. L14:1617 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 -1.82 and -1.731 values between -1.63 and -1.541 values between -1.35 and -1.252 values between -0.96 and -0.873 values between -0.87 and -0.772 values between -0.68 and -0.583 values between -0.58 and -0.4915 values between -0.49 and -0.3936 values between -0.39 and -0.30137 values between -0.30 and -0.20390 values between -0.20 and -0.11916 values between -0.11 and -0.011182 values between -0.01 and 0.08740 values between 0.08 and 0.18351 values between 0.18 and 0.2788 values between 0.27 and 0.3727 values between 0.37 and 0.467 values between 0.46 and 0.562 values between 0.56 and 0.651 values between 0.84 and 0.941 values between 1.22 and 1.321 values between 1.70 and 1.801 values between 1.89 and 1.99offnegativepositiveoften

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. raise
  2. etta
  3. 万人
  4. One
  5. Pix
  6. ra
  7. raise
  8. holder
  9. ruc
  10. 許

Suppresses

  1. aternity
  2. ides
  3. positor
  4. ackbar
  5. информацион
  6. sabot
  7. deactivated
  8. ilm
  9. 不清楚
  10. liberated