Neuron Explorer

Qwen3-0.6B

zeros inside long numbers

A neuron in layer 5 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 reacts to a zero digit in the middle of a long number, such as '2,500,000' or '0.0000000001'. Turning it up did not produce any clear change in the topic of the writing.

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

    0 meters or as 2570000000 millimeters.⏎- Colorado (western U.S.) 233

  2. #2 peak

    of a nanometer-sized magnetic grain in just a few picoseconds (1 picosecond is 0.000000000001 seconds). As a comparison, today’s hard drives take a few nanoseconds

  3. #3 peak

    0/15 IP: 10101100.00010000.00000000.00000000 SM: 1111

  4. #4 peak

    this child will become a doctor by the age of 40 years.⏎What I have just described<|endoftext|>|Area||2,500,000 sq km|⏎|Lowest Point||-5,267 m|⏎

  5. #5 peak

    drinking water after treatment with chlorinated compounds. Brines from shale gas wells contain dissolved bromide at concentrations up to 2,000,000 ppb and bromide can react with the chlorinated disinfectants and with

  6. #6 peak

    600,000 year earlier than had previously been documented.⏎- Our ancestors began making multicomponent

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 have a new friend, I have a new friend, and I have a new friend again. I have a new friend again, and I have a new friend again. So, I have four new friends. What is the total number

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

    0⏎network => 172.16.0.0/15 IP: 101011

  2. #2 peak

    aryons, which are hadrons made of three quarks, make up almost all the matter we see in the universe today," said Brookhaven theoretical physicist Swagato Mukherjee, a co-author on a paper describing the new results in

  3. #3 peak

    diverse career in live performance presentations includes: Entertainment Operations Manager for the Tropicana, Production Coordinator for Lincoln Center Festivals, and Associate Producer/Writer for Westinghouse TV. He currently serves on the Executive Board of the Stage Managers’ Association

  4. #4 peak

    the plant.⏎Young plants should be watered Every day.⏎7) Ripsali bacciferra-⏎8) string of tears-⏎Pack of 8 hanging succulent⏎About the combo :⏎Succulents are the plants that store

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. L2:1918 excites
  2. L4:1233 suppresses
  3. L1:1276 excites
  4. L1:21 suppresses
  5. L4:1180 suppresses
  6. L4:943 suppresses
  7. L2:1791 suppresses
  8. L3:219 weakly suppresses
  9. L4:1401 weakly suppresses unclear: letter fragments inside names
  10. L0:2319 weakly excites
  11. L0:2471 weakly excites
  12. L1:2211 weakly suppresses
  13. L1:1087 weakly suppresses
  14. L1:2288 weakly excites
  15. L1:1598 weakly excites
  16. L4:1117 weakly excites
  17. L0:1811 weakly suppresses
  18. L2:676 weakly suppresses
  19. L0:3043 weakly suppresses
  20. L1:1827 weakly suppresses
  21. L4:1808 weakly suppresses
  22. L4:1049 weakly excites
  23. L2:2442 weakly suppresses
  24. L0:1130 weakly suppresses

Feeds later layers

  1. L14:246 strongly suppresses
  2. L15:115 excites
  3. L17:166 suppresses
  4. L19:529 suppresses
  5. L19:1004 excites
  6. L17:360 excites
  7. L18:570 excites
  8. L20:83 suppresses
  9. L10:1228 excites
  10. L20:1040 suppresses
  11. L22:316 suppresses
  12. L15:96 suppresses
  13. L18:533 excites
  14. L10:445 suppresses
  15. L18:404 excites 'get', 'go' and 'got' after a subject
  16. L22:384 excites
  17. L18:137 suppresses 'the' after 'to' in fixed phrases
  18. L18:541 suppresses
  19. L18:288 excites
  20. L20:53 suppresses
  21. L17:389 excites
  22. L21:1112 suppresses
  23. L19:1865 suppresses
  24. L22:1141 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 -0.79 and -0.744 values between -0.64 and -0.594 values between -0.59 and -0.548 values between -0.54 and -0.4910 values between -0.49 and -0.4430 values between -0.44 and -0.3954 values between -0.39 and -0.3378 values between -0.33 and -0.28104 values between -0.28 and -0.23182 values between -0.23 and -0.18238 values between -0.18 and -0.13364 values between -0.13 and -0.08602 values between -0.08 and -0.031203 values between -0.03 and 0.02514 values between 0.02 and 0.07218 values between 0.07 and 0.1295 values between 0.12 and 0.1757 values between 0.17 and 0.2239 values between 0.22 and 0.2723 values between 0.27 and 0.3217 values between 0.32 and 0.3718 values between 0.37 and 0.4216 values between 0.42 and 0.4711 values between 0.47 and 0.523 values between 0.52 and 0.571 values between 0.57 and 0.625 values between 0.62 and 0.673 values between 0.67 and 0.721 values between 0.82 and 0.872 values between 0.87 and 0.921 values between 0.92 and 0.971 values between 1.07 and 1.121 values between 1.17 and 1.22offnegativepositiveoften

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. nd
  2. keiten
  3. es
  4. NING
  5. 肩
  6. ister
  7. loor
  8. ecute
  9. nings
  10. (pr

Suppresses

  1. ented
  2. orrow
  3. 行って
  4. Ỗ
  5. leased
  6. icator
  7. tees
  8. amarin
  9. 书法家
  10. returned