Neuron Explorer

Qwen3-0.6B

'or' linking two alternatives

A neuron in layer 16 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 'or' when it joins two alternative names for the same thing, such as 'boundary or interface', 'patella or the kneecap' or 'symbol or underlying'. It is most common in technical explanations. Turning it up or down did not produce a clear, related change in the writing.

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

    single ray of light strikes a boundary between two mediums and is refracted.⏎- The light ray which travels through the incident medium or first medium and strikes the boundary or interface is called the incident ray.⏎- The light ray which travels into the second

  2. #2 peak

    the vastus lateralis, vastus medialis, vastus intermedius, and the rectus femoris. These muscles are present just above the patella or the kneecap and form the patellar tendon. These muscles perform the function

  3. #3 peak

    options contract has the following properties or attributes:⏎- Expiration Date⏎- Symbol or Underlying⏎- Strike Price⏎- Type: Call or Put⏎The contract

  4. #4 peak

    <|endoftext|>Cancellous bone, also called trabecular bone or spongy bone, light, porous bone enclosing numerous large spaces that give a honeycombed or spongy appearance. The bone matrix, or framework, is organized into a

  5. #5 peak

    one is redundancy, so if one drive breaks down, the data will be accessible from the remaining ones, and the second is improved performance as the input/output, or reading/writing operations will be distributed among different drives. There are different RAID types

  6. #6 peak

    which represents the final prediction or output.⏎C. Splitting Data based on Features⏎The key feature of decision trees is their ability to split data based on features or

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 want to be able to have a small part of a small part of a small part of a small part of a small part of a small part of a small part of a small part of a small part of a small part of a

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

    The first self-evident truth of the Declaration of Independence says that “all men are created equal.” No words in the American lexicon are as eloquent or as powerful. We believe this to be a fundamental moral truth. Our Constitution begins

  2. #2 peak

    standard Teletype paper tape uses only 8 bits.) Cities would hire private companies to translate the data into reports that traffic engineers could use to adjust traffic lights or improve roads.⏎Bill Gates and Paul Allen were high school students at Lakeside School

  3. #3 peak

    short burst of time,, it can be overwhelming. This is because neurodivergent brains process information differently, making it hard for them to switch between tasks quickly or efficiently in stressful situations.⏎Neurodivergent individuals may also become distracted or

  4. #4 peak

    ⏎Machine Learning can be performed using multiple approaches. Even if you are not planning to become a machine learning developer, knowing these techniques can enable you how your current or future AI project will be built.⏎The four basic techniques of ML are: supervised

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. L15:3059 suppresses 'and' or 'or' joining the last list items
  2. L12:1443 excites opening parenthesis after a name or term
  3. L6:990 excites
  4. L10:135 excites 'or' in alternative requirements
  5. L15:727 excites
  6. L15:2935 suppresses
  7. L9:2010 suppresses
  8. L12:2076 excites
  9. L9:92 suppresses "and" or "or" joining two items
  10. L14:50 excites
  11. L10:144 suppresses
  12. L14:2276 excites opening bracket after a term or number
  13. L3:474 excites
  14. L15:515 suppresses word before 'that', 'to' or 'for' continuation
  15. L5:330 excites opening parenthesis after a noun
  16. L15:1907 excites the word 'or' between alternatives
  17. L10:1459 suppresses
  18. L9:1065 weakly suppresses
  19. L14:2603 weakly excites
  20. L15:693 weakly excites commas inside numbers and asides
  21. L12:2252 weakly suppresses
  22. L14:1209 weakly excites
  23. L11:406 weakly suppresses
  24. L10:1643 weakly suppresses

Feeds later layers

  1. L17:2274 excites
  2. L18:2544 excites
  3. L17:2704 excites
  4. L17:2258 excites
  5. L17:3025 excites
  6. L18:1916 excites
  7. L18:192 suppresses
  8. L19:2867 excites unclear: 'or something' and open parentheses
  9. L19:331 suppresses
  10. L18:933 excites
  11. L19:451 suppresses
  12. L19:1675 weakly excites
  13. L19:356 weakly suppresses
  14. L17:1214 weakly excites
  15. L21:1402 weakly excites
  16. L19:1438 weakly excites
  17. L20:1339 weakly excites
  18. L19:354 weakly suppresses
  19. L18:2136 weakly excites
  20. L17:1075 weakly suppresses
  21. L19:2711 weakly excites
  22. L17:667 weakly excites
  23. L20:2794 weakly excites
  24. L17:180 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.

83 values between -0.83 and -0.432114 values between -0.43 and -0.031530 values between -0.03 and 0.3756 values between 0.37 and 0.7734 values between 0.77 and 1.1715 values between 1.17 and 1.5818 values between 1.58 and 1.985 values between 1.98 and 2.388 values between 2.38 and 2.785 values between 2.78 and 3.185 values between 3.18 and 3.587 values between 3.58 and 3.996 values between 3.99 and 4.392 values between 4.39 and 4.793 values between 4.79 and 5.193 values between 5.59 and 5.992 values between 5.99 and 6.392 values between 6.39 and 6.803 values between 6.80 and 7.201 values between 7.20 and 7.601 values between 7.60 and 8.001 values between 8.40 and 8.801 values between 8.80 and 9.201 values between 10.81 and 11.211 values between 11.21 and 11.611 values between 14.83 and 15.23offnegativepositiveoften

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. itch
  2. ��
  3. yah
  4. hay
  5. phans
  6. công
  7. الماضي
  8. scratch
  9. prech
  10. insic

Suppresses

  1. ameleon
  2. 門
  3. 美的
  4. AdapterManager
  5. }}>⏎
  6. 门
  7. *>(
  8. ][/
  9. .SuspendLayout
  10. 花开