Neuron Explorer

Qwen3-0.6B

'can' after 'and' or a noun subject

A neuron in layer 21 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 word 'can' when it follows a subject that describes a possible result, as in 'holding illiquid assets can be a problem'. The opposite side fires on 'will' after 'you', and turning it up gave repeated 'be' loops.

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

    , a household, or a business, must have the proper balance between income producing assets and liquid assets such as cash which produce no income. Holding illiquid assets can be a problem when cash is quickly needed. The consequences can be having to sell

  2. #2 peak

    lead to surgical complications when your dog gets anesthetized.⏎- ACL injuries: The excess fat from obesity puts too much pressure on the ACL ligament and can result in ACL injuries, which can be very excruciating for your canine.⏎

  3. #3 peak

    how to report it. Furthermore, there is a lack of clear policies and procedures for dealing with sexual harassment in educational institutes. This creates confusion and uncertainty, and can make it difficult for victims to seek help.⏎Another major issue is the lack of

  4. #4 peak

    .⏎Alternative orthodontic treatments have long been available but may not always be as comprehensive as orthodontics. Space maintainers help maintain space for adult teeth and can prevent complications and the need for more orthodontic therapy. Removable computer-generated

  5. #5 peak

    concerns about the potential for PET bottles to release harmful chemicals when they are disposed of. Like all plastics, PET bottles are made from non-renewable resources and can take hundreds of

  6. #6 peak

    . This makes it difficult for businesses to forecast the cost of fuel and can cast doubts over the future of their supply chain.⏎• Maintenance: Due to its

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 best way in the world is (s) ) (s) ) (s) ) (s) ) (s) ) (s) ) (s) ) (s) ) (s) ) (s) ) (

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

    problem?⏎Our tutors have indicated that to solve this problem you will need to apply the Conjugate Acids and Bases concept. If you

  2. #2 peak

    Mobile App: Apple Store | Google Play⏎Dear visitor, the comments section is undergoing an overhaul and will return soon.<|endoftext|>|You might also like:||Lotto (English)||Lotto Game (English): Board #

  3. #3 peak

    AdobeID at the Adobe website. For more details please see FAQ 1&2. To view this ebook on an iPhone, iPad or Android mobile device you will need the Adobe Digital Editions app, or BlueFire Reader or Txtr app

  4. #4 peak

    cesis.⏎For Nietzsche, destiny keeps the form of Amor fati (Love of Fate) through the important element of Nietzsche's philosophy, the "will to power" (der Wille zur Macht), the basis of human behavior,

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. L20:636 strongly suppresses
  2. L19:1678 strongly excites
  3. L19:1208 suppresses
  4. L18:2303 suppresses
  5. L20:2089 suppresses
  6. L19:545 excites
  7. L19:3044 suppresses
  8. L20:2802 suppresses
  9. L19:1769 excites
  10. L18:736 excites
  11. L20:2524 excites
  12. L17:1297 excites
  13. L19:730 excites
  14. L19:858 suppresses
  15. L15:886 excites
  16. L19:2156 suppresses
  17. L19:1633 excites
  18. L10:2166 excites
  19. L20:2527 excites
  20. L20:2262 excites
  21. L19:2491 suppresses
  22. L17:1994 suppresses
  23. L17:953 excites the word 'can' after a subject
  24. L18:2917 suppresses

Feeds later layers

  1. L24:298 strongly excites 'would' in hypothetical or past statements
  2. L25:364 strongly excites
  3. L24:314 strongly suppresses
  4. L22:627 strongly excites
  5. L24:864 strongly excites 'to' before an infinitive
  6. L27:247 strongly excites
  7. L26:296 strongly excites
  8. L26:225 strongly excites 'and' joining two abstract nouns
  9. L27:401 strongly excites
  10. L23:257 strongly suppresses
  11. L27:201 strongly excites
  12. L26:862 strongly excites
  13. L25:503 strongly excites
  14. L26:2432 strongly excites
  15. L27:515 strongly excites
  16. L26:675 strongly excites
  17. L22:479 strongly suppresses
  18. L27:880 strongly excites
  19. L26:671 strongly excites
  20. L25:600 strongly excites
  21. L27:559 strongly excites
  22. L27:963 excites
  23. L27:900 excites
  24. L26:769 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 -19.73 and -18.521 values between -12.45 and -11.241 values between -11.24 and -10.022 values between -10.02 and -8.812 values between -8.81 and -7.594 values between -7.59 and -6.382 values between -6.38 and -5.172 values between -5.17 and -3.954 values between -3.95 and -2.7434 values between -2.74 and -1.522864 values between -1.52 and -0.31784 values between -0.31 and 0.9148 values between 0.91 and 2.1224 values between 2.12 and 3.3320 values between 3.33 and 4.5519 values between 4.55 and 5.7618 values between 5.76 and 6.9817 values between 6.98 and 8.1915 values between 8.19 and 9.404 values between 9.40 and 10.6210 values between 10.62 and 11.839 values between 11.83 and 13.054 values between 13.05 and 14.265 values between 14.26 and 15.473 values between 15.47 and 16.694 values between 16.69 and 17.901 values between 17.90 and 19.121 values between 19.12 and 20.331 values between 20.33 and 21.543 values between 22.76 and 23.971 values between 27.61 and 28.83offnegativepositiveoften

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. be
  2. ensure
  3. быть
  4. determine
  5. produce
  6. execute
  7. define
  8. create
  9. make
  10. specify

Suppresses

  1. 开了
  2. 带来了
  3. 笑了
  4. 获得了
  5. 掉了
  6. 来了
  7. 进入了
  8. 打开了
  9. 选择了
  10. 提升了