Neuron Explorer

Qwen3-0.6B

inside numbers and citations

A neuron in layer 16 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the pieces of numbers, dates, and reference details, such as a digit in a postcode, a dot in an author initial, or the dollar sign before a sum. It goes negative on dashes in year ranges and commas in plant names. Turning it up made the model drift into number lists.

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

    references: Gwartney, J. D., Stroup, R. L., Sobel, R. S., & Macpherson, D. A. (2018). Economics: Private⏎and public choice (16

  2. #2 peak

    Water-Quality Assessment (NAWQA) Program of the U.S. Geological Survey (USGS).1 Chloroform and other trihalomethanes, the most commonly detected compounds, were found in about 9 percent of the sampled

  3. #3 peak

    570 and at Hammocks Beach State Park located at 1572 State Rd 1511, Swansboro, NC 28584.⏎The Coastal Federation’s Recycling for Reefs Program collects shells and

  4. #4 peak

    in reality is brown, colors were produced by putting together thousands of of NASA’s images of Mercury surface, Space Environment, Geochemistry, and Ranging (Known as Messenger) spacecraft, which is in orbit around the innermost planet. Engineers used

  5. #5 peak

    2025 and 2030. Respectively, these targets are 15% and 37.5% lower than the 2021 figure. The targets were finalized in 2019

  6. #6 peak

    and flows generally southwestward, watering the Paris Basin, to join the Seine River at Conflans, after a course of 188 miles (302 km). It traverses the Thiérache Hills, where it collects

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 number of a few minutes, the number of a few minutes, the number of a few minutes, the number of a few minutes, the number of a few minutes, and, in some way, the number of a few minutes,

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

    , including handling of certain kinds of global variables.⏎The first step in creating a GLib shared library is to create a ``static'' library (also called an ar archive).

  2. #2 peak

    results⏎Authors: Walker, Robert J. (John), 1801-1869⏎Date: May 27, 185

  3. #3 peak

    for interior growing conditions. It also likes to be potbound and will bloom even in a relatively tiny pot.⏎Common name: Fire lily, bush lily, clivia⏎Botanical name: Clivia miniata⏎Flowers: Rounded

  4. #4 peak

    and examples of With clause in oracle let us first check some important bullet points of With Clause :⏎- With Clause in Oracle is released in Oracle 9i release 2 to improve the performance of complex sql queries.⏎- The clause works like a

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. L15:7 excites
  2. L14:597 suppresses unclear: document ends and list breaks
  3. L12:1014 suppresses
  4. L13:2302 suppresses
  5. L15:73 weakly excites
  6. L8:1219 weakly suppresses
  7. L15:370 weakly suppresses
  8. L14:1915 weakly suppresses
  9. L12:730 weakly suppresses
  10. L13:1345 weakly suppresses
  11. L10:1731 weakly excites
  12. L5:84 weakly excites
  13. L5:75 weakly suppresses
  14. L8:177 weakly excites
  15. L3:533 weakly suppresses
  16. L2:49 weakly excites
  17. L2:1537 weakly suppresses
  18. L11:1231 weakly suppresses
  19. L5:71 weakly suppresses
  20. L8:379 weakly excites
  21. L10:1700 weakly excites
  22. L7:1265 weakly suppresses
  23. L12:229 weakly suppresses
  24. L4:1647 weakly excites

Feeds later layers

  1. L19:2731 weakly excites
  2. L23:1772 weakly suppresses
  3. L20:1417 weakly excites
  4. L21:1724 weakly suppresses
  5. L18:2235 weakly suppresses
  6. L19:752 weakly suppresses
  7. L20:1448 weakly excites "the" in mid-sentence, weak
  8. L27:1605 weakly excites
  9. L17:2646 weakly suppresses
  10. L17:1088 weakly excites
  11. L22:1349 weakly excites
  12. L20:2133 weakly suppresses letters inside spelled-out words
  13. L19:1925 weakly excites
  14. L17:1397 weakly suppresses
  15. L19:2225 weakly suppresses
  16. L20:2510 weakly suppresses
  17. L19:2030 weakly suppresses
  18. L27:235 weakly suppresses
  19. L24:183 weakly excites
  20. L23:2817 weakly excites the word "no" in "no longer"
  21. L18:2347 weakly excites
  22. L17:497 weakly suppresses
  23. L22:621 weakly excites
  24. L23:247 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.

1 values between -3.44 and -3.232 values between -3.02 and -2.824 values between -2.82 and -2.616 values between -2.61 and -2.409 values between -2.40 and -2.1912 values between -2.19 and -1.9816 values between -1.98 and -1.7820 values between -1.78 and -1.5744 values between -1.57 and -1.3647 values between -1.36 and -1.1584 values between -1.15 and -0.94116 values between -0.94 and -0.74173 values between -0.74 and -0.53251 values between -0.53 and -0.32333 values between -0.32 and -0.11540 values between -0.11 and 0.10518 values between 0.10 and 0.30448 values between 0.30 and 0.51344 values between 0.51 and 0.72253 values between 0.72 and 0.93202 values between 0.93 and 1.14151 values between 1.14 and 1.34101 values between 1.34 and 1.5571 values between 1.55 and 1.7648 values between 1.76 and 1.9723 values between 1.97 and 2.1819 values between 2.18 and 2.3815 values between 2.38 and 2.5916 values between 2.59 and 2.808 values between 2.80 and 3.016 values between 3.01 and 3.2212 values between 3.22 and 3.425 values between 3.42 and 3.632 values between 3.63 and 3.842 values between 3.84 and 4.053 values between 4.05 and 4.261 values between 4.26 and 4.462 values between 4.67 and 4.88offnegativepositiveoften

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. Strauss
  2. idar
  3. NavController
  4. tin
  5. stabilization
  6. arity
  7. /*!<
  8. +:+
  9. interfaces
  10. strt

Suppresses

  1. OMET
  2. ount
  3. omet
  4. adox
  5. leg
  6. 簿
  7. oles
  8. 忘了
  9. ちゃんと
  10. 울