Neuron Explorer

Qwen3-0.6B

'of' in 'piece of' phrases

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 'of' in phrases like 'a piece of' or 'pieces of', where a small part of something is named. The negative side is weak and not clearly meaningful. Turning it up did not produce a clear theme in the sample text.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    munch on tender soft new growth. Like many other animals, deer have food they avoid, and<|endoftext|>Seven miles to the south west of Sheffield there's a piece of wild moorland surrounded by roads. It's called Big Moor. I estimate

  2. #2 peak

    motor oil will help to extend the life of your vehicle’s engine and make sure everything performs to specifications.⏎On every bottle of motor oil, there are three pieces of information: the API service rating, the viscosity grade and the “Energy Conserving

  3. #3 peak

    System for Interactive Scientific and Scientific-Technical Information”.<|endoftext|>An team of researchers heading to the Antarctic later this year may go in search of a long-lost piece of exploration history. A group of international climate scientists will travel into the Southern Ocean to

  4. #4 peak

    when he was a kid. All you have to do is find a piece of paper and use your imagination to draw any prehistoric scene you’d like. But

  5. #5 peak

    is Multi-Factor Authentication (MFA)?⏎Multi-factor authentication is a mechanism whereby a user can only access a website or application after successfully presenting two or more pieces of evidence that indicates they are

  6. #6 peak

    flowing through the circuit exceeds a specified limit, thus preventing damage to the electrical components.⏎In the industrial and industry context, a fixture refers to a device or piece of equipment that is used to securely hold, support, or position a workpiece during

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: and the we, and the 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

    ass boilers; Bubbling fluidized bed boilers; Circulating fluidized bed boilers⏎- Front<|endoftext|>CNY - 中國元,人民幣⏎最常用 CNY 交叉匯率⏎為什麼您

  2. #2 peak

    the beginning, explaining the history of Java and why it's⏎important. You then learn how to obtain the Java Development Kit (JDK)⏎and write your first Java program.⏎Next, it's on to the Java fundamentals, including data types

  3. #3 peak

    be. Her. Self.”⏎Have you been in the presence of someone who holds space for you when you’re talking? I can think of one friend who does this so well – my friend Danielle.⏎When we’re sitting

  4. #4 peak

    in order to seek better conditions and pay for the railroad workers. The most famous of the ARU strikes was the Pullman strike, or Debs Rebellion⏎2

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. L13:44 suppresses inside web addresses and citations
  2. L10:155 excites
  3. L15:184 excites
  4. L15:59 suppresses
  5. L13:1320 suppresses
  6. L4:2589 suppresses
  7. L8:1673 suppresses
  8. L14:836 suppresses
  9. L8:328 excites
  10. L5:7 suppresses
  11. L7:92 excites
  12. L2:24 suppresses
  13. L5:125 weakly excites "of" after all three, both, ten
  14. L11:223 weakly excites
  15. L8:16 weakly suppresses
  16. L13:608 weakly suppresses opening parenthesis starting an aside
  17. L14:2680 weakly excites
  18. L14:1832 weakly suppresses
  19. L2:2533 weakly excites
  20. L1:1818 weakly excites
  21. L12:158 weakly suppresses
  22. L15:1554 weakly excites
  23. L6:1009 weakly excites
  24. L14:1929 weakly suppresses

Feeds later layers

  1. L19:2712 suppresses
  2. L18:2608 excites
  3. L21:274 suppresses
  4. L19:272 suppresses
  5. L25:1005 weakly excites
  6. L22:576 weakly suppresses
  7. L17:258 weakly excites
  8. L18:973 weakly suppresses
  9. L19:663 weakly excites
  10. L20:519 weakly suppresses
  11. L25:2408 weakly excites
  12. L17:1467 weakly suppresses
  13. L19:2930 weakly suppresses
  14. L21:716 weakly excites
  15. L18:2303 weakly suppresses
  16. L18:1642 weakly suppresses
  17. L25:109 weakly excites
  18. L22:194 weakly suppresses
  19. L27:905 weakly suppresses
  20. L19:914 weakly suppresses
  21. L21:3031 weakly excites
  22. L18:3023 weakly suppresses
  23. L17:985 weakly excites
  24. L21:511 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.

2 values between -0.96 and -0.7913 values between -0.79 and -0.6354 values between -0.63 and -0.47212 values between -0.47 and -0.31608 values between -0.31 and -0.141244 values between -0.14 and 0.021151 values between 0.02 and 0.19450 values between 0.19 and 0.35106 values between 0.35 and 0.5135 values between 0.51 and 0.675 values between 0.67 and 0.843 values between 0.84 and 1.003 values between 1.00 and 1.164 values between 1.16 and 1.331 values between 1.33 and 1.492 values between 1.49 and 1.651 values between 1.82 and 1.983 values between 1.98 and 2.142 values between 2.14 and 2.311 values between 2.47 and 2.631 values between 2.63 and 2.801 values between 2.96 and 3.121 values between 3.29 and 3.453 values between 3.45 and 3.612 values between 5.41 and 5.57offnegativepositiveoften

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. ty
  2. 或多
  3. side
  4. nde
  5. اسب
  6. hy
  7. ensions
  8. 新的
  9. qui
  10. loo

Suppresses

  1. 的一个
  2. 的第一个
  3. 了一遍
  4. 了一口
  5. 一句
  6. 了一会
  7. 了个
  8. 一枚
  9. 一件
  10. 了一场