Neuron Explorer

Qwen3-0.6B

word before a noun naming people or a role

A neuron in layer 10 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 fires on the word just before a noun that names a group of people or a role, such as 'six crews', 'people that coach' or 'people with disabilities'. Turning it up did not make the written text show a clear pattern.

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

    to be launched to test the Kontakt lunar rendezvous/docking system. During the Salyut 1 mission, in June 1971, six crews were identified and assigned to specific Soyuz spacecraft serial numbers for these tests. These

  2. #2 peak

    to another client, think twice before parting with your economic rights.<|endoftext|>The USA Basketball Player Development Curriculum has been established to guide players, and the people that coach them, through a level-appropriate system of basketball development. Using scientific guiding principles

  3. #3 peak

    and put his scalp on display for money.⏎In 1900, Thompson sent the scalp back to the Nebraska doctor who tried to reaffix it, who, in turn, donated it to the public library. The scalp remained on display

  4. #4 peak

    Porcelain⏎Process – Raw Porcelain is rolled to create the tile, each letter is hand imprinted into the clay, which is then fired in the potter’s kiln at 1230 degrees celsius, making it solid

  5. #5 peak

    a moon base, a representative wouldn’t confirm or deny the plans. “We are pursuing a range of possible destinations on route to an eventual trip to Mars,” they explained, but admitted that a lunar waypoint would be “in the range of possibilities

  6. #6 peak

    Each season, natural<|endoftext|>This brochure is one response to a need identified by people with disabilities. The way we portray people with disabilities and our attitudes toward them are critical

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 am 25 years old, and I have been thinking about it." This is the case of the movie "The Granger" (2005) who is 25, and the movie "The Granger

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

    their hands. As they develop control of their body, they may become hyperactive. Always on the move, these students must find physically active careers⏎Logical – those who excel in problem-solving & number manipulation. They may be easily frustrated with anything

  2. #2 peak

    glance, a relatively simple question with a quick answer: Hindus speak for Hindus. If proof is needed, then one could commission a poll to ask a cross-section of Hindus who speaks for them. But a series of articles on this very question appearing

  3. #3 peak

    are called to be servants. Scriptures seem to invite that identity and it makes sense: Love your neighbor as yourself. Serve others as Jesus served. Let the greatest among you be a servant.⏎Remember, though, that many biblical images for “serv

  4. #4 peak

    Since ECT is well-known for having negative effects on recall in general — that’s why it’s reserved for only the most intractable cases— the researchers wanted to see if it might only disrupt a

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. L9:1248 strongly excites
  2. L9:782 strongly suppresses
  3. L9:205 suppresses
  4. L7:360 suppresses
  5. L9:273 suppresses
  6. L9:3023 suppresses
  7. L4:885 excites
  8. L7:180 excites
  9. L9:1948 excites
  10. L7:743 suppresses
  11. L8:1751 excites
  12. L7:2714 excites
  13. L6:214 suppresses
  14. L7:2753 suppresses
  15. L9:1562 excites 'by' before a byline or agent
  16. L5:17 suppresses
  17. L6:699 excites
  18. L9:1221 suppresses
  19. L9:1830 excites
  20. L9:1652 suppresses
  21. L6:563 excites
  22. L6:2756 excites
  23. L5:2728 excites
  24. L6:682 excites

Feeds later layers

  1. L16:1054 suppresses
  2. L15:1244 suppresses
  3. L11:1741 suppresses
  4. L16:2215 suppresses
  5. L26:1294 suppresses
  6. L12:495 suppresses
  7. L16:1438 suppresses
  8. L12:2415 suppresses
  9. L12:1679 suppresses
  10. L22:1997 suppresses
  11. L11:333 suppresses
  12. L13:2765 suppresses
  13. L12:1479 suppresses
  14. L12:2887 suppresses
  15. L11:1167 suppresses
  16. L11:1512 suppresses
  17. L22:396 suppresses
  18. L22:545 suppresses
  19. L11:978 suppresses
  20. L16:1894 excites
  21. L16:2627 suppresses
  22. L12:755 suppresses
  23. L11:1844 suppresses
  24. L11:1211 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 -5.84 and -5.571 values between -5.29 and -5.021 values between -4.48 and -4.211 values between -4.21 and -3.942 values between -3.94 and -3.672 values between -3.67 and -3.401 values between -3.12 and -2.855 values between -2.85 and -2.587 values between -2.58 and -2.319 values between -2.31 and -2.0412 values between -2.04 and -1.7718 values between -1.77 and -1.5023 values between -1.50 and -1.2343 values between -1.23 and -0.9552 values between -0.95 and -0.68101 values between -0.68 and -0.41170 values between -0.41 and -0.141065 values between -0.14 and 0.132025 values between 0.13 and 0.40287 values between 0.40 and 0.6719 values between 0.67 and 0.9418 values between 0.94 and 1.227 values between 1.22 and 1.4913 values between 1.49 and 1.766 values between 1.76 and 2.035 values between 2.03 and 2.303 values between 2.30 and 2.572 values between 2.57 and 2.843 values between 2.84 and 3.113 values between 3.11 and 3.391 values between 3.39 and 3.661 values between 3.66 and 3.931 values between 4.74 and 5.01offnegativepositiveoften

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. erable
  2. ibur
  3. itemap
  4. Magnum
  5. rake
  6. forks
  7. ikal
  8. Jarvis
  9. rary
  10. metast

Suppresses

  1. who
  2. who
  3. ptide
  4. whom
  5. 에게
  6. 群体
  7. 人群中
  8. 的目光
  9. 目光
  10. 一人