Neuron Explorer

Qwen3-0.6B

'and' splitting a two-part category

A neuron in layer 7 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

The neuron lights up on the word 'and' (or a comma) when a text is splitting something into two or three named kinds, such as 'simple and complex carbohydrates' or 'land and aquatic habitats'. Turning it up made the model's text drift toward unrelated topics, so the steering result doesn't show a clear effect.

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

    monosaccharides (oligosaccharides and polysaccharides).⏎From a chemical point of view, it makes sense to divide carbohydrates into simple and complex. However, this division is not enough to explain what happens to the different

  2. #2 peak

    necessities for life such as shelter, water, and food for organisms to live, grow, and reproduce. The earth has two basic types of habitats: land habitats and aquatic habitats. On land, there are dry to wet forests (deserts to

  3. #3 peak

    review the steps you can take to help protect yourself further.⏎How Does Encryption Work?⏎Encryption is based on a key that has two different parts. The public part and the private part. The public part of the key is distributed to those you

  4. #4 peak

    .). Hudson's Bay Co. still administers Rupert's Land. Louisiana is ceded to Spain by France.⏎Following the Constitutional Act, Quebec is divided into Upper and Lower Canada (1791). Spain cedes

  5. #5 peak

    (also known as blood sugar) and used by the cells for energy purposes.⏎What are complex carbohydrates?⏎Carbohydrates are divided into two main categories: simple and complex carbohydrates. Simple carbohydrates (monosaccharides and disaccharides

  6. #6 peak

    For example, pressure will be needed when creating deep indents.⏎Lathes are divided into three types: the engine lathe, special purpose lathes and turret lathes. Some variants are placed on benches and others more portable.

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 add a function to a function I am trying to add a function to a function I am trying to add a function to a function I am trying to add a function to a function I am trying to add

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

    ⏎B. Classful Model⏎C. Classless Model⏎D. Heterogeneous Model⏎6-5. Which of the following is correct in CIDR?⏎A. Class A includes Class B network⏎B. There are only two

  2. #2 peak

    -chained peripheral devices simultaneously. The latest USB 2.0 specification can deliver 480Mbps data transfer bandwidth. In addition, USB provides plug-and-play capabilities to allow device changes while the computer is powered on.

  3. #3 peak

    Python, R, and Scala so that you start creating your Spark projects and collaborating with other data scientists. When you sign up, you get free access to Watson Studio and all other IBM services for 30 days. Start

  4. #4 peak

    light source can carry light over mass distances, ranging from a few inches to over 100 miles. There are two kinds of fiber optics. The single-mode fiber optic is used for high speed and long distance transmissions because they have extremely tiny

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. L0:1118 weakly suppresses
  2. L1:668 weakly excites
  3. L5:2834 weakly suppresses
  4. L6:2017 weakly excites
  5. L5:236 weakly excites
  6. L2:2032 weakly suppresses
  7. L5:738 weakly excites
  8. L5:1748 weakly suppresses
  9. L6:2036 weakly excites
  10. L1:1960 weakly excites
  11. L0:1669 weakly suppresses
  12. L0:890 weakly suppresses
  13. L2:1712 weakly suppresses
  14. L6:1113 weakly excites
  15. L2:2887 weakly suppresses
  16. L0:2635 weakly suppresses
  17. L6:1626 weakly suppresses
  18. L6:897 weakly excites
  19. L6:1947 weakly suppresses
  20. L4:1037 weakly excites
  21. L6:2451 weakly suppresses
  22. L4:1498 weakly suppresses
  23. L3:2904 weakly suppresses
  24. L1:1364 weakly excites

Feeds later layers

  1. L10:1186 excites
  2. L8:193 suppresses
  3. L14:1957 excites
  4. L8:877 excites
  5. L13:2441 excites
  6. L10:1964 excites
  7. L11:2453 excites
  8. L10:2449 excites
  9. L13:86 excites
  10. L12:574 excites
  11. L12:2840 excites
  12. L13:16 excites
  13. L11:773 excites
  14. L8:986 weakly excites
  15. L10:2501 weakly excites
  16. L9:2980 weakly excites
  17. L12:1731 weakly suppresses
  18. L8:2999 weakly excites
  19. L10:782 weakly excites
  20. L12:2987 weakly suppresses
  21. L9:1852 weakly excites
  22. L15:336 weakly suppresses
  23. L8:1226 weakly excites
  24. L8:343 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 -2.17 and -2.062 values between -0.90 and -0.801 values between -0.80 and -0.692 values between -0.69 and -0.594 values between -0.59 and -0.4827 values between -0.48 and -0.3889 values between -0.38 and -0.27302 values between -0.27 and -0.17759 values between -0.17 and -0.061512 values between -0.06 and 0.04717 values between 0.04 and 0.15312 values between 0.15 and 0.25108 values between 0.25 and 0.3644 values between 0.36 and 0.476 values between 0.47 and 0.574 values between 0.57 and 0.683 values between 0.68 and 0.781 values between 0.78 and 0.894 values between 0.89 and 0.993 values between 0.99 and 1.103 values between 1.10 and 1.201 values between 1.41 and 1.521 values between 1.52 and 1.621 values between 1.62 and 1.731 values between 1.94 and 2.05offnegativepositiveoften

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. ayer
  2. iden
  3. ldr
  4. ayers
  5. caption
  6. 传统产业
  7. ernen
  8. ekt
  9. ErrorMessage
  10. iggs

Suppresses

  1. mäßig
  2. }/#{
  3. RH
  4. acing
  5. 樣
  6. uring
  7. curring
  8. 调剂
  9. 管理水平
  10. Lansing