Neuron Explorer

Qwen3-0.6B

digit 1 inside dates and times

A neuron in layer 25 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 responds to the number 1 when it appears inside a date or a clock time, such as "2012/10/05" or "12:30". When it was turned down, the model got stuck writing long strings of the digit 3.

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

    impact on the community in the Hollywood Gazette’s November 2019 issue — https://hollywoodgazette.com/2019/11/03/hollywood-history-the-heroes-and-the-history-of-liberia

  2. #2 peak

    |⏎|rs_class_altemira.tif||30.7 MB||1985-01-01 - 2001-12-31||2.50S To 4.00

  3. #3 peak

    as lasting? Some believe not.⏎Taken from…. The Future of Education: Tablets vs. Textbooks http://mashable.com/2012/10/05/tablets-vs-textbooks/⏎They also say…. ‘ Students

  4. #4 peak

    413967-1-768x146-1-300x57.png admin2021-10-20 23:34:59202

  5. #5 peak

    and a show. After you eat peanut butter & jelly, watch Stone Soup, performed by the Lost Caravan. Lunch is at noon; show starts at 12:30. Chesapeake Music Hall, Off Rt.

  6. #6 peak

    to know before planting⏎- Article by: Deb Brown⏎- Special to the Star Tribune⏎- January 27, 2009 - 12:32 PM⏎Q It seems like birch trees grow well around

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: 33333333333333333333333333333333333333333333333

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

    in hospital equipment such medical equipment such as centrifuges and automated pipetting systems.⏎One of the most common applications of lead screws is for positioning the print head in 3D printers. They often use 8mm lead screws as demonstrated in the

  2. #2 peak

    laser scanning.⏎In the overall Scan-to-BIM process, a laser scanner is used to acquire a precise 3D scan of the project’s real-world conditions.⏎The scan data is then

  3. #3 peak

    tumour right when it starts and know its next move, maybe we can also stop it from growing at all.⏎EMBL scientists have created a new, realistic 3D testbed that could help them achieve that goal by studying cancer cells as

  4. #4 peak

    the area of the cavity. At worst, small holes or channels are completely blocked.⏎Printing with light: DLP, SLA, LCD⏎Modern resin-based 3D printing works on the principle of photopolymerization. In other words

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. L22:1641 excites
  2. L22:2983 excites
  3. L22:934 suppresses
  4. L22:747 excites
  5. L20:815 suppresses
  6. L10:825 weakly suppresses 'B' in vitamin B names
  7. L22:1474 weakly excites
  8. L7:2676 weakly suppresses
  9. L24:1968 weakly excites
  10. L11:1255 weakly excites
  11. L0:1549 weakly suppresses
  12. L6:2795 weakly excites
  13. L5:2103 weakly excites
  14. L24:2412 weakly excites
  15. L14:1991 weakly excites
  16. L23:231 weakly excites
  17. L4:423 weakly suppresses
  18. L17:1228 weakly suppresses
  19. L23:1695 weakly excites
  20. L23:1166 weakly excites
  21. L3:2123 weakly suppresses
  22. L15:2377 weakly excites
  23. L1:2989 weakly excites
  24. L18:1517 weakly suppresses

Feeds later layers

  1. L26:463 strongly suppresses digits inside years and numbers
  2. L26:102 strongly excites
  3. L27:364 strongly suppresses
  4. L27:555 strongly suppresses
  5. L26:724 strongly suppresses
  6. L26:983 strongly suppresses
  7. L27:678 suppresses
  8. L26:667 suppresses
  9. L27:287 suppresses
  10. L27:113 excites
  11. L27:132 excites
  12. L27:2172 excites
  13. L27:992 suppresses
  14. L27:448 excites
  15. L26:476 suppresses
  16. L27:696 suppresses
  17. L27:163 excites
  18. L26:2953 excites
  19. L27:428 suppresses
  20. L26:611 suppresses
  21. L27:1093 suppresses
  22. L26:1083 suppresses
  23. L26:270 suppresses
  24. L26:1127 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 -13.08 and -12.501 values between -11.35 and -10.771 values between -10.77 and -10.202 values between -8.47 and -7.893 values between -7.89 and -7.311 values between -7.31 and -6.742 values between -6.74 and -6.162 values between -6.16 and -5.592 values between -5.59 and -5.011 values between -5.01 and -4.432 values between -4.43 and -3.869 values between -3.86 and -3.282 values between -3.28 and -2.705 values between -2.70 and -2.1314 values between -2.13 and -1.5529 values between -1.55 and -0.97195 values between -0.97 and -0.401947 values between -0.40 and 0.181431 values between 0.18 and 0.76232 values between 0.76 and 1.3315 values between 1.33 and 1.911 values between 1.91 and 2.482 values between 2.48 and 3.061 values between 3.06 and 3.641 values between 3.64 and 4.213 values between 4.21 and 4.791 values between 4.79 and 5.371 values between 8.25 and 8.821 values between 9.40 and 9.98offnegativepositiveoften

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. .centerY
  2. zept
  3. elerik
  4. ubb
  5. veyor
  6. ypi
  7. ้อง
  8. alis
  9. Intl
  10. $($

Suppresses

  1. 3
  2. three
  3. three
  4. trois
  5. Three
  6. três
  7. 3
  8. THREE
  9. -three
  10. 三个