Neuron Explorer

Qwen3-0.6B

comma inside large numbers

A neuron in layer 3 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 a comma sitting inside a big number, like the comma in "$157,000" or "6,000 lbs". Turning it up made the writing repetitive and looping, which does not show much about the comma itself.

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

    Lubantuun in 1927.”⏎The Crystal Skull is often referred to as the “Skull of Doom”; the skull was made from a single, unusually large block of clear

  2. #2 peak

    Curtiss and purchase a plane for $7,000 ($157,000 in today’s money). While waiting for it to be built,

  3. #3 peak

    4 wheels required to carry a 6,000-lb load would use the following calculation:⏎[(4,000 lbs + 6,000 lbs) × 1.25] ÷ 4

  4. #4 peak

    14 Star Tribune<|endoftext|>For practical purposes, the M1 Ultra acts like a single, impossibly large slice of silicon that does it all. Apple’s most powerful chip

  5. #5 peak

    of effectiveness) each year produces over $350,000 more income for his/her students compared to an average teacher. On

  6. #6 peak

    the commands you give and the M-file are somewhat disparate. The process is illustrated using a single, simple differential equation:⏎Integrate this equation from t = 0 to t =

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 "b" and "c" in the "b" and "c" market. It is in the "b" and "c" market. It is in the "b" and "c" market. It is in

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

    into what the Framers found important in 1781, and what they changed their minds on by 1788.⏎They might have been onto something here:⏎Term limit for legislative office⏎Articles: No more than three out

  2. #2 peak

    remote-controlled torpedoes and devices that had applications in missile defense and computers, long after his death in 1943.⏎Now the race is on to preserve Tesla's<|endoftext|>For years, the feeding habits and nutritional requirements of the b

  3. #3 peak

    , is stranger: they became symbols.⏎There is a single species of palm native to the entire state of California, the California fan palm, which is a big one with what looks like a fuzzy beard of brown leaves underneath its green fronds.

  4. #4 peak

    Fisher (pictured above) and his band of treasure hunters finally found the big one. ‘Nuestra Señora de Atocha’ was the largest and

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. L2:1718 suppresses
  2. L0:2376 weakly excites
  3. L2:2123 weakly excites
  4. L2:49 weakly suppresses
  5. L1:1464 weakly excites
  6. L2:2169 weakly excites
  7. L0:1981 weakly suppresses
  8. L1:2832 weakly suppresses
  9. L0:1301 weakly suppresses
  10. L2:111 weakly suppresses
  11. L2:552 weakly excites comma inside a stacked adjective list
  12. L0:879 weakly excites
  13. L2:788 weakly excites
  14. L2:1657 weakly excites
  15. L0:2320 weakly excites
  16. L2:960 weakly suppresses
  17. L2:92 weakly excites
  18. L2:2810 weakly suppresses
  19. L1:1522 weakly suppresses
  20. L2:1628 weakly excites
  21. L0:1504 weakly suppresses
  22. L1:2922 weakly excites
  23. L1:788 weakly suppresses
  24. L0:239 weakly suppresses numbers and dates with stray punctuation

Feeds later layers

  1. L5:2799 excites
  2. L4:146 excites
  3. L8:1871 weakly suppresses
  4. L6:2086 weakly excites
  5. L6:2164 weakly excites second half of multi-word proper names
  6. L4:1867 weakly excites
  7. L17:2874 weakly suppresses
  8. L22:844 weakly suppresses
  9. L15:1165 weakly excites
  10. L22:903 weakly suppresses
  11. L10:1010 weakly suppresses 'in the process of' and continuing actions
  12. L26:2462 weakly suppresses
  13. L20:1293 weakly suppresses
  14. L4:1106 weakly suppresses
  15. L18:1463 weakly suppresses
  16. L4:245 weakly suppresses
  17. L17:184 weakly excites
  18. L13:1035 weakly suppresses
  19. L14:42 weakly excites
  20. L4:160 weakly excites
  21. L23:2556 weakly suppresses
  22. L7:351 weakly excites
  23. L15:1964 weakly excites
  24. L8:2718 weakly excites

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.

2 values between -1.01 and -0.801 values between -0.80 and -0.5912 values between -0.59 and -0.39163 values between -0.39 and -0.181857 values between -0.18 and 0.021619 values between 0.02 and 0.23208 values between 0.23 and 0.4419 values between 0.44 and 0.647 values between 0.64 and 0.854 values between 0.85 and 1.054 values between 1.05 and 1.262 values between 1.47 and 1.671 values between 1.67 and 1.881 values between 2.08 and 2.291 values between 2.50 and 2.701 values between 2.91 and 3.111 values between 3.32 and 3.521 values between 3.52 and 3.731 values between 5.79 and 6.001 values between 6.00 and 6.201 values between 6.20 and 6.411 values between 7.03 and 7.23offnegativepositiveoften

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. 亲
  2. ème
  3. 请选择
  4. przy
  5. ыта
  6. éd
  7. avel
  8. .Cho
  9. êt
  10. ฉ

Suppresses

  1. igure
  2. 参与到
  3. opyright
  4. 几何
  5. rup
  6. ('('
  7. rush
  8. 같이
  9. wagon
  10. stärke