Neuron Explorer

Qwen3-0.6B

verb right after a long subject

A neuron in layer 17 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 fire on the verb ('are', 'is', 'has') that comes right after a long noun phrase subject, such as 'Live tutors are' or 'Optimization capabilities and limits of modern compilers are'. Turning it up did not make a clear, related change in the text.

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

    generation of ELF object files, and static and dynamic linking. Code examples and exercises are included along with the best code practices. Optimization capabilities and limits of modern compilers are examined, enabling you to balance between program readability and performance. The use of various

  2. #2 peak

    Price Calculator⏎With this bond price calculator, we aim to help you calculate the bond price issued by a government or a corporation. Finding out the current bond price is one of the most critical procedures for bond investors, as miscalculating can

  3. #3 peak

    the book. Crucially, a broad scope of organic chemistry is covered – which is as it should be. Familiar topics, such as drug discovery and pesticides are explained extremely well. Importantly

  4. #4 peak

    a little to get valuable, extensive worksheets that actually assist your child to learn.⏎Worksheets that have been developed by a preschool instructor that knows what the school demands are and also what your youngster requires to learn. A collection of worksheets that cover all

  5. #5 peak

    this provides critical feedback to teachers about how well students are performing in the concepts taught by them. If teachers find that most students are going wrong in a particular concept, they can arrange for remedial classes for such concepts and help students gain proficiency in

  6. #6 peak

    met with one of this alert messages already. Or, even worse, users reached out to you saying your website or eCommerce store isn’t secure according to Google. And, correct me if I’m wrong, you were caught on short, someone might

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: it's the first time in a month, the day of a month day of the month day of a month day of the month, and the day of a month day of the month day of the month day of the month day of 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

    ing the states of Indiana, Ohio, Illinois, and Michigan. Convenient access to raw materials, skilled labor, and the shipping options available including the railroads, likely contributed to the concentration of these companies within this area. Another factor was a receptive

  2. #2 peak

    informed choices regarding their studies and future careers.⏎The essence of SLB⏎The importance of SLB is becoming increasingly apparent. The job market is changing and this has significant implications for future workers (current students). Occupations are changing content faster than

  3. #3 peak

    oliath grouper. Additionally, the goliath grouper were falling victim to overfishing, red tide, and changes in climate. All of these factors prompted FWC to ban the killing of this species. To date, the policy

  4. #4 peak

    of both parties in recent years.⏎And at the legislative level, the focus will be on suburban districts like those in Polk County, where Democrats hope booming population will lead to more favorable districts. Population loss in some rural areas could also mean

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. L16:2942 suppresses
  2. L16:2557 excites
  3. L14:2641 suppresses
  4. L16:2924 suppresses
  5. L16:2710 excites
  6. L14:2944 suppresses
  7. L14:1284 excites
  8. L15:2585 excites
  9. L7:1234 suppresses
  10. L6:2820 suppresses
  11. L16:2654 excites
  12. L14:1158 excites 'to' starting a purpose clause
  13. L11:1085 excites
  14. L13:1815 weakly suppresses
  15. L16:2236 weakly suppresses
  16. L6:2610 weakly suppresses
  17. L16:411 weakly excites
  18. L15:2366 weakly excites
  19. L16:1332 weakly excites
  20. L6:904 weakly suppresses
  21. L15:1543 weakly suppresses
  22. L4:1120 weakly excites
  23. L12:2200 weakly suppresses
  24. L7:2289 weakly excites

Feeds later layers

  1. L18:391 suppresses
  2. L19:1608 suppresses
  3. L18:727 excites
  4. L20:1093 suppresses
  5. L20:813 suppresses
  6. L21:943 suppresses
  7. L18:1744 weakly excites
  8. L19:2180 weakly suppresses
  9. L20:578 weakly suppresses
  10. L24:194 weakly suppresses
  11. L19:617 weakly suppresses
  12. L20:457 weakly suppresses
  13. L26:436 weakly suppresses
  14. L26:669 weakly suppresses
  15. L19:640 weakly suppresses
  16. L27:626 weakly suppresses
  17. L19:806 weakly suppresses
  18. L18:69 weakly suppresses
  19. L19:2123 weakly excites
  20. L18:1235 weakly suppresses
  21. L19:2888 weakly excites
  22. L19:2932 weakly suppresses
  23. L24:852 weakly excites
  24. L21:596 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 -13.63 and -13.151 values between -12.20 and -11.723 values between -10.30 and -9.821 values between -9.82 and -9.351 values between -9.35 and -8.871 values between -8.40 and -7.922 values between -7.92 and -7.454 values between -7.45 and -6.972 values between -6.97 and -6.501 values between -6.50 and -6.025 values between -6.02 and -5.553 values between -5.55 and -5.074 values between -5.07 and -4.593 values between -4.59 and -4.124 values between -4.12 and -3.645 values between -3.64 and -3.179 values between -3.17 and -2.699 values between -2.69 and -2.2213 values between -2.22 and -1.7415 values between -1.74 and -1.2746 values between -1.27 and -0.79335 values between -0.79 and -0.321952 values between -0.32 and 0.16924 values between 0.16 and 0.63314 values between 0.63 and 1.11130 values between 1.11 and 1.5864 values between 1.58 and 2.0631 values between 2.06 and 2.5411 values between 2.54 and 3.016 values between 3.01 and 3.493 values between 3.49 and 3.962 values between 3.96 and 4.441 values between 4.44 and 4.912 values between 4.91 and 5.39offnegativepositiveoften

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. lij
  2. 没有想到
  3. isée
  4. 得不到
  5. 大家分享
  6. 能得到
  7. 都无法
  8. 리
  9. 没想到
  10. .removeEventListener

Suppresses

  1. rings
  2. ib
  3. screens
  4. (ViewGroup
  5. -cloud
  6. class
  7. edu
  8. urma
  9. cstdint
  10. Crop