Neuron Explorer

Qwen3-0.6B

'in' after success or help verbs

A neuron in layer 1 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 responds to the word 'in' when it comes right after verbs such as 'succeed', 'assist' or 'help', as in 'succeed in doing'. It is pushed strongly negative on 'in' that links a place or container, like 'stored in files'. Turning it up made the output odd and repetitive, which does not clearly match the trigger.

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

    voluntary, non-dinding process in which an impartial and neutral mediator assists the parties in reaching an agreement. A mediator does not impose a solution but creates conducive environment in

  2. #2 peak

    to prevent them from effecting their object. Some delay occurred before the brigade began its march; and the Commander-in-Chief, seeing that the Russians would succeed in carrying off the guns if not at once attacked, despatched Captain Nolan,

  3. #3 peak

    write drew public attention, put him under the limelight, and made him the only loud voice that had tremendously increase his credibility. In addition, Douglass succeeds in using a chronological flow of narrative that clearly recounts his previous miserable life condition under which

  4. #4 peak

    to be well-informed citizens.<|endoftext|>"You're a strong skater, Piet, and you have a quick mind. This is why I know you'll succeed in this important task. I wouldn't ask you to do this if I didn't

  5. #5 peak

    and uncover the remarkable ways in which they assist us in diagnosing and managing our health.⏎Table of Contents⏎Cancer is a disease

  6. #6 peak

    . At the urging of fellow scientists he attempted to form an imperial academy of science, but was thwarted by political enemies of the Jesuits. He did succeed in publishing a very timely and indispensable journal concerning the latest scientific discoveries. A lunar crater

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 need to be able to do with a way to get the information and I need to be able to do with a way to get the information. So, I have to go to the information and I have to go to the information.

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

    ACTS tool captures trace links automatically. The trace links are captured in an XML file.⏎Scalability, readability issues with XML file, links are stored locally in files therefore not accessible to all users.⏎The goal of my project is to aid

  2. #2 peak

    viewable from any Internet - enabled computer in the world.⏎Files are just everywhere - they are the 'building material' of the World Wide Web, storing data in the form of text and numbers, which makes it possible for us to obtain any

  3. #3 peak

    configuration sections that a developer might define. It then looks in the application configuration file. To keep the machine configuration file manageable, it is best to put these settings in the application configuration file. However, putting the settings in the machine configuration file can

  4. #4 peak

    used to downloading data from the Cloud onto mobile devices, most do not know where the data comes from. Digital information is stored in minute magnetic dots written in layers that are only a few nanometers thick that cover

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. L0:743 weakly suppresses
  2. L0:114 weakly excites
  3. L0:1519 weakly excites
  4. L0:2819 weakly suppresses
  5. L0:310 weakly excites
  6. L0:3033 weakly suppresses
  7. L0:1154 weakly suppresses
  8. L0:1280 weakly excites
  9. L0:1137 weakly suppresses
  10. L0:463 weakly excites
  11. L0:269 weakly suppresses
  12. L0:652 weakly excites
  13. L0:754 weakly suppresses
  14. L0:2121 weakly suppresses
  15. L0:1455 weakly excites
  16. L0:794 weakly excites
  17. L0:634 weakly excites
  18. L0:2347 weakly suppresses
  19. L0:1641 weakly excites
  20. L0:1132 weakly excites
  21. L0:2415 weakly suppresses
  22. L0:189 weakly suppresses
  23. L0:2633 weakly excites
  24. L0:2611 weakly excites split-off 's' starting a word

Feeds later layers

  1. L11:1573 weakly excites
  2. L22:1351 weakly suppresses
  3. L20:2314 weakly excites
  4. L20:2676 weakly excites
  5. L18:1985 weakly excites
  6. L15:1727 weakly excites
  7. L17:1767 weakly excites
  8. L6:1858 weakly excites
  9. L8:525 weakly excites
  10. L17:2133 weakly suppresses
  11. L14:2023 weakly suppresses
  12. L17:2011 weakly excites
  13. L20:241 weakly excites
  14. L23:310 weakly excites
  15. L27:434 weakly excites
  16. L22:53 weakly suppresses
  17. L23:988 weakly excites
  18. L6:2626 weakly excites
  19. L11:334 weakly suppresses
  20. L4:913 weakly excites
  21. L19:539 weakly excites
  22. L19:1590 weakly excites
  23. L18:616 weakly excites
  24. L21:1502 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 -2.27 and -2.162 values between -2.04 and -1.931 values between -1.93 and -1.821 values between -1.59 and -1.482 values between -1.37 and -1.262 values between -1.26 and -1.142 values between -1.14 and -1.033 values between -1.03 and -0.927 values between -0.92 and -0.815 values between -0.81 and -0.691 values between -0.69 and -0.5810 values between -0.58 and -0.4714 values between -0.47 and -0.3660 values between -0.36 and -0.24330 values between -0.24 and -0.131245 values between -0.13 and -0.021752 values between -0.02 and 0.10373 values between 0.10 and 0.2149 values between 0.21 and 0.3220 values between 0.32 and 0.436 values between 0.43 and 0.555 values between 0.55 and 0.664 values between 0.66 and 0.774 values between 0.77 and 0.882 values between 0.88 and 1.002 values between 1.11 and 1.223 values between 1.22 and 1.331 values between 1.45 and 1.561 values between 2.12 and 2.23offnegativepositiveoften

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. 企业提供
  2. ilate
  3. .HORIZONTAL
  4. Soph
  5. 说了
  6. Traffic
  7. ivities
  8. лом
  9. уть
  10. ions

Suppresses

  1. ATUS
  2. acas
  3. _ASSERT
  4. aser
  5. 芜
  6. das
  7. dbg
  8. vốn
  9. alist
  10. .setdefault