Neuron Explorer

Qwen3-0.6B

zeros after a decimal point

A neuron in layer 24 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 reacts to the zero in small decimal numbers like 0.01, 0.2 or 0.003, the zeros that come right after the decimal point. When it was turned up, the model often began writing long strings of zeros.

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

    to it being the basis of every single financial transaction.⏎In today’s American coinage, there are six circulating denominations. These are the penny ($0.01

  2. #2 peak

    also sometimes called step size) to determine the next point. For example, if the gradient magnitude is 2.5 and the learning rate is 0.01, then the gradient descent algorithm will pick the next point 0.0

  3. #3 peak

    here to learn more about weed control and weed and feed.⏎Clay soil is comprised of small, tightly packed particles with a diameter less than 0.002mm. If you examine the structure of these particles, you’ll find it

  4. #4 peak

    will make a microscope slide, or thin section, of their rock of interest. A thin section has a very thin slice of rock, usually about 0.03 mm thick, attached to a glass slide with an epoxy resin. Thin sections

  5. #5 peak

    by the Modelica Association.⏎This example demonstrates different noise distributions methods that can be selected for a Noise block. Both noise blocks use samplePeriod = 0.02 s, y_min=-1, y_max=3, and have identical

  6. #6 peak

    breadth and depth remain the most extensive, unexplored habit.⏎Activity 2 – Freshwater Habitats⏎Freshwater Habitats cover less than 0.01 % of the earth’s

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: the state of a game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game game

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 National Hurricane Center can keep tabs on them.⏎Once a cyclone’s winds kick up to 39 miles per hour and sustain that speed for 10 minutes, it becomes a tropical storm and the NHC gives it a name (

  2. #2 peak

    2, 144, 180, and 200, resulting in stepping angles of 30, 15, 5, 2.5, 2, and 1.8° per

  3. #3 peak

    a straight line. The reason the path is all bendy is because Swift itself is in motion! Swift orbits the Earth, circling us once every 90 minutes or so. As it moves around us, its viewpoint is changing, and

  4. #4 peak

    - Hernias: Hernia is the popular term for a series of different diagnoses, all identified by outpouchings on the ventral aspect of the abdominal wall of pigs. Hernias can be painful and limit

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. L23:2500 excites
  2. L19:843 suppresses
  3. L21:3057 suppresses
  4. L20:626 excites
  5. L4:498 weakly suppresses
  6. L20:2749 weakly suppresses
  7. L23:1324 weakly suppresses
  8. L10:1052 weakly suppresses
  9. L3:1124 weakly suppresses
  10. L10:726 weakly suppresses
  11. L9:977 weakly suppresses the last digit of a decade year
  12. L20:1504 weakly excites
  13. L16:184 weakly suppresses
  14. L7:2244 weakly suppresses
  15. L9:344 weakly suppresses
  16. L2:976 weakly suppresses
  17. L5:935 weakly suppresses
  18. L10:934 weakly suppresses
  19. L12:1587 weakly excites
  20. L6:2792 weakly excites
  21. L21:203 weakly excites
  22. L0:997 weakly suppresses
  23. L13:2294 weakly suppresses
  24. L1:3053 weakly excites

Feeds later layers

  1. L25:473 excites zeros inside large round numbers
  2. L26:413 excites
  3. L27:1153 excites
  4. L27:862 excites
  5. L27:1196 excites
  6. L27:153 suppresses the spot just before a historical number
  7. L27:396 excites
  8. L27:530 suppresses
  9. L25:182 weakly excites
  10. L25:92 weakly excites
  11. L26:121 weakly excites
  12. L26:737 weakly excites
  13. L27:1111 weakly excites
  14. L26:682 weakly suppresses
  15. L25:212 weakly excites
  16. L25:543 weakly suppresses
  17. L25:1308 weakly excites
  18. L25:324 weakly suppresses
  19. L26:233 weakly suppresses
  20. L25:2243 weakly suppresses
  21. L25:2022 weakly excites
  22. L27:628 weakly suppresses
  23. L25:1895 weakly excites first digit of a year like 2020
  24. L27:1029 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 -13.56 and -12.521 values between -12.52 and -11.482 values between -10.45 and -9.411 values between -9.41 and -8.371 values between -8.37 and -7.331 values between -7.33 and -6.291 values between -6.29 and -5.258 values between -5.25 and -4.214 values between -4.21 and -3.1710 values between -3.17 and -2.1326 values between -2.13 and -1.092090 values between -1.09 and -0.061669 values between -0.06 and 0.9846 values between 0.98 and 2.0212 values between 2.02 and 3.067 values between 3.06 and 4.104 values between 4.10 and 5.146 values between 5.14 and 6.185 values between 6.18 and 7.223 values between 7.22 and 8.261 values between 8.26 and 9.301 values between 9.30 and 10.342 values between 10.34 and 11.382 values between 11.38 and 12.411 values between 12.41 and 13.451 values between 18.65 and 19.692 values between 26.96 and 28.00offnegativepositiveoften

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. 0
  2. 零
  3. zero
  4. zero
  5. htag
  6. ₀
  7. 0
  8. Zero
  9. heets
  10. zept

Suppresses

  1. 驷
  2. ystate
  3. <State
  4. fine
  5. affen
  6. ucid
  7. 该游戏
  8. 衫
  9. loses
  10. 輩