Neuron Explorer

Qwen3-0.6B

'and' joining two time or quantity ranges

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 "and" when it joins two things of the same kind, especially times, periods or amounts, like "late winter and continuing" or "26 states and D.C.". The strongest negative cases are dashes or "to" in number ranges. Turning it up did not give a clear topic.

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

    feeding, starting in late winter and continuing until the ice breaks up.”⏎Pacific walrus use the floating pack ice both

  2. #2 peak

    this is the result of too much air and not enough gasoline getting mixed together in the internal combustion chamber. The “lean” word refers to the abundance of air and not enough fuel.⏎Meanwhile, there is very little gasoline being burnt

  3. #3 peak

    , the result is corruption.⏎In The Great Gatsby, F. Scott Fitzgerald illustrates the American Dream and its corruption through the evolution of a society from the wealth and social statuses they achieved. One way that Fitzgerald illustrates the

  4. #4 peak

    anical name: Clivia miniata⏎Flowers: Rounded clusters of red, orange or yellow trumpet-shaped flowers appear on tall, thick stalks in late winter and early spring, lasting for several weeks.⏎Foliage:

  5. #5 peak

    9 new astronauts in April 1966.⏎Pogue’s first and only space assignment was as Command Module Pilot for Skylab 4, the third and final manned flight to the Skylab space station. He and Commander Gerald Carr and

  6. #6 peak

    changed a lot in the past century. From individually packaged food servings to disposable diapers, more garbage is generated now than ever before. The average American discards seven and a half pounds of garbage every day. This garbage, the solid waste stream,

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: thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about thinking about

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

    in an Education Program Session today at 7:30 a.m. and again at 2:00 p.m. in rooms 393-396 of the Ernest N. Morial Convention Center. The session is

  2. #2 peak

    heat gain. Tight construction, good insulation and high-performance windows are key. "Weather stripping and caulking in existing homes can reduce your energy bill 10 to 25 percent," says David Johnston, president of What's Working, a

  3. #3 peak

    inea pig at home.⏎As rodents, guinea pigs are considered large weighing approximately 700 and 1200 grams or 1.5 to 2.5 pounds and measuring around 20 to

  4. #4 peak

    , whole grains, nuts and yogurt.⏎This is based on following 120,000 people over 20 years. They were normal weight to

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. L15:727 excites
  2. L14:2202 excites
  3. L14:2865 suppresses
  4. L9:1449 suppresses
  5. L14:706 excites
  6. L12:1786 excites
  7. L7:1762 excites
  8. L15:1501 excites "and" joining two parts of a name or phrase
  9. L15:852 suppresses
  10. L9:92 suppresses "and" or "or" joining two items
  11. L5:1469 suppresses
  12. L6:1706 suppresses
  13. L15:2722 excites "and" or "or" in a pair or list
  14. L9:2010 suppresses
  15. L15:545 suppresses
  16. L3:351 weakly suppresses
  17. L15:628 weakly suppresses
  18. L15:564 weakly excites
  19. L15:1922 weakly excites
  20. L13:2356 weakly excites
  21. L15:2178 weakly excites
  22. L8:1083 weakly excites "x" or "by" between two dimensions
  23. L15:752 weakly excites
  24. L12:2548 weakly excites

Feeds later layers

  1. L17:465 suppresses
  2. L17:595 suppresses
  3. L17:1575 suppresses
  4. L18:1625 suppresses
  5. L18:1413 suppresses
  6. L17:1214 suppresses
  7. L17:1438 suppresses
  8. L17:878 suppresses
  9. L18:2136 suppresses
  10. L17:2516 suppresses
  11. L19:2562 suppresses
  12. L18:1386 suppresses unclear: words before 'and' or 'or'
  13. L17:1965 suppresses
  14. L17:3025 suppresses
  15. L18:2354 suppresses
  16. L19:1849 suppresses
  17. L17:1092 suppresses
  18. L17:2274 suppresses
  19. L20:516 suppresses
  20. L17:1498 suppresses
  21. L17:1679 suppresses
  22. L18:1572 weakly suppresses
  23. L17:1259 weakly suppresses
  24. L19:1823 weakly suppresses 'and' inside organisation or paired names

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 -13.21 and -12.701 values between -12.18 and -11.671 values between -8.58 and -8.061 values between -7.55 and -7.031 values between -6.52 and -6.004 values between -3.95 and -3.431 values between -3.43 and -2.923 values between -2.92 and -2.407 values between -2.40 and -1.896 values between -1.89 and -1.3719 values between -1.37 and -0.8653 values between -0.86 and -0.342679 values between -0.34 and 0.171035 values between 0.17 and 0.6942 values between 0.69 and 1.2021 values between 1.20 and 1.7212 values between 1.72 and 2.234 values between 2.23 and 2.753 values between 2.75 and 3.265 values between 3.26 and 3.782 values between 3.78 and 4.291 values between 4.29 and 4.812 values between 4.81 and 5.322 values between 5.32 and 5.841 values between 6.35 and 6.861 values between 6.86 and 7.38offnegativepositiveoften

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. ongan
  2. inal
  3. (Attribute
  4. __________________⏎⏎
  5. conc
  6. 廓
  7. \System
  8. *I
  9. 年人
  10. (IR

Suppresses

  1. .forChild
  2. ebx
  3. hev
  4. .MoveNext
  5. 其余
  6. 末
  7. ัว
  8. арь
  9. abella
  10. 절