Neuron Explorer

Qwen3-0.6B

timestamps and dates in text

A neuron in layer 19 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 reacts to the spot right after a date or time, such as the gap before the hour in 'October 3rd, 2010, 07:44'. Steering it did not produce a clear theme.

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

    Internet (<|endoftext|>このトピックには0件の返信が含まれ、1人の参加者がいます。2020年7月6日 at 9:45 AMに Megan Smith さんが最後の更新を行いました。⏎

  2. #2 peak

    AdminStaff1 · Views: 1,130⏎Contributed by: AdminStaff1 · April 10, 2008 @ 12

  3. #3 peak

    .com⏎Fri Sep 10 00:10:22 CEST 2010⏎On 9/9/10 4:39 PM, Baba wrote:⏎> In below code "

  4. #4 peak

    will print you need to store it in 6 bytes, with the last byte equal to 0.⏎October 3rd, 2010, 07:44 AM⏎We are not going to do you homework.⏎

  5. #5 peak

    7.png admin2021-10-20 23:34:592021-10-20 23:34:59assignment 2⏎mynursing

  6. #6 peak

    . Fonts used: school script dashed, Rudiment and Homemade Apple.⏎Posted by Craftyminx on October 12, 2011 at

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 collection of time and other global cultural and national cultural products of my country in my work in my office in some of my local and international. I have some of my own local and international national and regional products. Now,

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

    and 48% were influenza A(H3N2).⏎(FluNet data as of 9 August 2011 12:50 UTC)⏎* Some NICs report to FluNet retrospectively leading to updates of

  2. #2 peak

    much smaller than a human. Essential oils are highly concentrated and should<|endoftext|>Monday, October 5, 1998 Published at 17:47 GMT

  3. #3 peak

    'Lose my life, my leg or my hand?': Modern-day slavery in India⏎10 July 2014 Last updated at 18:48 BST⏎It is a form of modern day slavery otherwise known as forced

  4. #4 peak

    3(5):2947-2952.<|endoftext|>- Published on Tuesday, 18 September 2007 20:10⏎- Written

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. L5:55 strongly suppresses
  2. L9:383 strongly excites
  3. L6:605 suppresses
  4. L17:923 suppresses
  5. L10:637 suppresses
  6. L9:1057 excites
  7. L7:2505 suppresses
  8. L4:2251 excites
  9. L8:1257 suppresses
  10. L7:2318 excites
  11. L4:1939 suppresses
  12. L18:1511 suppresses
  13. L16:583 excites
  14. L16:1668 suppresses
  15. L6:2289 suppresses
  16. L14:1088 excites
  17. L12:2338 excites
  18. L10:986 excites
  19. L3:3062 excites
  20. L4:2497 suppresses
  21. L6:2811 suppresses
  22. L8:2342 suppresses
  23. L4:2887 excites
  24. L2:2166 suppresses 'in' after high or low

Feeds later layers

  1. L20:2665 suppresses
  2. L22:1074 suppresses
  3. L20:1335 suppresses
  4. L20:442 suppresses
  5. L21:1185 suppresses
  6. L27:934 suppresses
  7. L25:565 suppresses
  8. L24:906 suppresses
  9. L24:2671 suppresses
  10. L24:2325 suppresses
  11. L22:2445 suppresses
  12. L22:504 excites
  13. L26:2567 suppresses
  14. L21:2852 weakly suppresses spot where a number or 'time' comes next
  15. L21:375 weakly excites
  16. L20:1288 weakly suppresses
  17. L22:1113 weakly suppresses
  18. L23:2882 weakly suppresses
  19. L21:1209 weakly suppresses
  20. L23:1205 weakly suppresses
  21. L20:1199 weakly suppresses
  22. L20:1680 weakly suppresses
  23. L22:2729 weakly suppresses
  24. L20:2347 weakly 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 -44.97 and -43.301 values between -41.63 and -39.961 values between -13.24 and -11.573 values between -11.57 and -9.901 values between -9.90 and -8.234 values between -8.23 and -6.5610 values between -6.56 and -4.8911 values between -4.89 and -3.2216 values between -3.22 and -1.55772 values between -1.55 and 0.123069 values between 0.12 and 1.799 values between 1.79 and 3.462 values between 3.46 and 5.133 values between 5.13 and 6.801 values between 8.47 and 10.141 values between 11.81 and 13.481 values between 15.15 and 16.821 values between 16.82 and 18.491 values between 20.16 and 21.83offnegativepositiveoften

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. -sur
  2. 水量
  3. aire
  4. NavItem
  5. Navigate
  6. ail
  7. ียง
  8. ثار
  9. filesize
  10. NAV

Suppresses

  1. Uhr
  2. .sleep
  3. UTC
  4. clock
  5. OfDay
  6. utc
  7. .Timer
  8. =time
  9. 틱
  10. 时间