Neuron Explorer

Qwen3-0.6B

comma inside numbers and dates

A neuron in layer 18 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 comma in written numbers and dates, such as the comma in 100,000 or in 'January 29, 2013'. Turning it up made the model's writing drift into strings of counting and repeated number-like text.

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

    .S. trials before the Nuremberg Military Tribunals (NMT) took place from December 9, 1946 to April 13, 1949

  2. #2 peak

    of the mother continent between 100,000 and 60,000<|endoftext|>Know Your Limits,

  3. #3 peak

    anoma skin cancer) to estimate the number of people in the U.S. who had ever been diagnosed with cancer and who were alive on Jan. 1, 2007.⏎As of that day, 65% of

  4. #4 peak

    potential investors and vie for prizes as large as $200,000 in value.⏎Helion Energy Fusion Engine has received about $7

  5. #5 peak

    our commitment to continued energy efficiency gains, sustainability, and lower operating costs for consumers and enterprises.⏎The study further shows that an enterprise customer using 100,000 PCs running 6th Generation AMD A-Series APU could save

  6. #6 peak

    required a great deal of structural analysis to ensure it can withstand seismic forces—and curious visitors.⏎January 29, 2013—It’s plain to see that turning the remains of

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: (1) (The) (1) (The) (1) (The) (1) (The) (1) (The) (1) (The) (1) (The) (1) (The)

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

    CO2 soil concentrations are due to the volcanic rumblings of the magma still active below Mammoth Mountain. At more subtle levels this research has been able to map, using hyperspectral air borne imagery, the tree plant stress over all of the Mam

  2. #2 peak

    a month later was life-changing for not just the Soviet Union and the US but for everyone on this planet,” Zelibor said. “It showed us all, we do not need to be bound by Earth’s gravity because we have an entire

  3. #3 peak

    war:⏎“Think, too, of the great part that is played by the unpredictable in war,” Thucydides wrote, “and think of it now, before you are actually committed. The longer a war lasts, the more things tend

  4. #4 peak

    the sacred isle of Columba! Here landing, he ascended the loftiest of the hills upon the isle, and "gazing into the distance, found no longer any trace

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. L14:35 strongly suppresses
  2. L16:240 strongly excites comma after 'sometimes' or 'the truth'
  3. L17:118 strongly excites comma after a clause ending
  4. L9:77 strongly excites commas in citations and name lists
  5. L4:3 strongly excites commas after citation fields
  6. L5:16 strongly excites
  7. L11:154 suppresses commas after citation numbers
  8. L10:335 excites commas in catalog, date, and list metadata
  9. L17:585 excites
  10. L2:66 excites
  11. L3:71 suppresses
  12. L15:630 suppresses
  13. L17:1405 suppresses
  14. L17:1597 excites
  15. L15:1213 suppresses
  16. L16:241 excites
  17. L16:846 excites comma opening an aside or explanation
  18. L15:588 excites comma after an introductory phrase
  19. L12:1557 suppresses comma after an introductory phrase
  20. L17:116 excites
  21. L1:86 excites
  22. L16:1939 excites line break mid-sentence in wrapped text
  23. L17:2221 weakly suppresses
  24. L17:2001 weakly excites

Feeds later layers

  1. L19:726 strongly excites comma inside large numbers
  2. L19:682 strongly suppresses
  3. L21:700 excites comma before a new clause
  4. L20:476 excites commas that lead into a continuation
  5. L25:880 excites comma after a cause or effect clause
  6. L19:2645 excites commas in the middle of a sentence
  7. L26:236 suppresses
  8. L20:2082 suppresses
  9. L23:184 excites
  10. L19:383 excites commas after a clause or list item
  11. L22:144 excites
  12. L27:86 excites
  13. L21:632 suppresses
  14. L22:368 suppresses
  15. L19:1036 suppresses
  16. L20:513 excites
  17. L21:2616 excites
  18. L19:2403 excites comma after a full clause
  19. L27:358 suppresses
  20. L19:2175 excites
  21. L25:328 suppresses
  22. L20:811 suppresses
  23. L20:173 suppresses
  24. L21:530 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 -6.61 and -5.741 values between -2.28 and -1.421408 values between -1.42 and -0.552250 values between -0.55 and 0.3138 values between 0.31 and 1.1827 values between 1.18 and 2.0412 values between 2.04 and 2.9114 values between 2.91 and 3.7715 values between 3.77 and 4.6413 values between 4.64 and 5.509 values between 5.50 and 6.3714 values between 6.37 and 7.239 values between 7.23 and 8.1014 values between 8.10 and 8.976 values between 8.97 and 9.8310 values between 9.83 and 10.7013 values between 10.70 and 11.5611 values between 11.56 and 12.436 values between 12.43 and 13.296 values between 13.29 and 14.165 values between 14.16 and 15.029 values between 15.02 and 15.891 values between 15.89 and 16.753 values between 17.62 and 18.483 values between 18.48 and 19.352 values between 19.35 and 20.212 values between 20.21 and 21.081 values between 21.08 and 21.942 values between 21.94 and 22.811 values between 22.81 and 23.672 values between 27.14 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. especially
  2. which
  3. StringComparison
  4. Inc
  5. YTE
  6. noting
  7. Cor
  8. którzy
  9. but
  10. including

Suppresses

  1. ,
  2. ,⏎⏎⏎⏎
  3. ,⏎⏎⏎
  4. 不下
  5. ,e
  6. ustin
  7. ,{⏎
  8. igraph
  9. ,...⏎
  10. ,request