Neuron Explorer

Qwen3-0.6B

decade and century numbers like 1950s

A neuron in layer 11 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 digits of years written as decades or centuries, such as the '9' in '1950s' or the gap before '17th century'. When it was turned up or down, the model's text did not visibly change in a clear way.

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

    first time in Italy in 1590. However, until the first decade of the 17TH century, there is no effective evidence of the birth of the telescope

  2. #2 peak

    970s bought those semi-enclosed pedestal-style pay phones which, for most of us (Superman included) found lacking.⏎By the 1990s there were nearly three million pay phones in America. But now just

  3. #3 peak

    just wasn’t much technological or scientific progress at that time.⏎What about the first decade of the 20th century? That’s easier. The Zeppelin airship. The

  4. #4 peak

    positive impacts on Puerto Rico. What major change hit Puerto Rico in the 1990s that would drastically hurt their economy?⏎By the 1950s, the flow of Puerto Ricans to the mainland United States had

  5. #5 peak

    NATO planning posited early or initial use of nuclear weapons like the 1964 Warsaw Pact plan from the Czech archives. Certainly, by the 1950s, NATO war plans assumed early use of nuclear weapons, even immediate

  6. #6 peak

    . Although this is a wonderful rifle that It also progressed through World War II, when most of the military relied on old bolt action rifles, by the 1950s the M1 was simply obsolete. By this time,

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 year 1996.50 1985 1945. I 65 1900.00. 60. 1900 1910

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

    of a Living Trust⏎Sometimes it becomes necessary to petition a court to have a trust terminated. In the State of Texas, Sec. 112.054 of the Texas Property Code governs judicial<|endoftext|>Adjustment to college life

  2. #2 peak

    explore the Orange Revolution. In this Orange Revolution lesson, students examine the climate of the Ukrainian protest and investigate the role of Canadian government in the protest.⏎3 Views 2 Downloads<|endoftext|>The environment contains those factors ‘surrounding the organization’.

  3. #3 peak

    -earth wall within the property of Boisbuchet⏎EUR 858: Students⏎EUR 1110: Professionals<|endoftext|>Created on Thursday, 03 July 2014 Written by THE BELLEFONTAINE

  4. #4 peak

    1700’s in order to win a prize, that the problem was solved.⏎Joan Marie Galat, Pelican ©2012, 978-1-4556-1637-

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. L6:1469 excites
  2. L10:868 suppresses
  3. L10:2461 excites
  4. L10:220 suppresses hyphen inside compound words
  5. L7:229 excites first digit of a split number
  6. L9:13 excites
  7. L4:973 suppresses
  8. L6:756 suppresses
  9. L6:714 excites
  10. L8:2933 weakly suppresses
  11. L6:1888 weakly suppresses
  12. L6:1149 weakly suppresses
  13. L6:641 weakly suppresses
  14. L8:2644 weakly excites
  15. L8:853 weakly excites slot right before a quantity
  16. L9:438 weakly suppresses
  17. L5:2113 weakly suppresses
  18. L2:1668 weakly suppresses
  19. L2:1582 weakly excites
  20. L4:53 weakly excites blank slot before a number
  21. L4:2510 weakly suppresses
  22. L0:1745 weakly suppresses quantity word after large, high or low
  23. L10:781 weakly excites
  24. L8:885 weakly excites

Feeds later layers

  1. L19:298 suppresses
  2. L21:175 suppresses
  3. L16:372 suppresses
  4. L18:291 excites
  5. L15:828 suppresses
  6. L15:2665 weakly suppresses
  7. L20:95 weakly excites
  8. L16:1569 weakly suppresses
  9. L20:382 weakly excites
  10. L13:1251 weakly excites
  11. L16:256 weakly suppresses number right after 'age of' or 'ages'
  12. L14:177 weakly excites
  13. L25:46 weakly excites
  14. L19:1428 weakly excites
  15. L16:434 weakly excites
  16. L18:991 weakly suppresses
  17. L12:2853 weakly suppresses
  18. L18:2897 weakly excites
  19. L19:2113 weakly excites
  20. L14:2082 weakly excites
  21. L25:209 weakly suppresses
  22. L14:1700 weakly suppresses
  23. L14:1690 weakly excites
  24. L19:245 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 -5.18 and -4.971 values between -3.09 and -2.882 values between -2.88 and -2.671 values between -2.67 and -2.462 values between -2.25 and -2.044 values between -2.04 and -1.842 values between -1.84 and -1.636 values between -1.63 and -1.429 values between -1.42 and -1.2112 values between -1.21 and -1.0021 values between -1.00 and -0.7931 values between -0.79 and -0.5872 values between -0.58 and -0.37222 values between -0.37 and -0.172040 values between -0.17 and 0.041197 values between 0.04 and 0.25204 values between 0.25 and 0.4627 values between 0.46 and 0.6715 values between 0.67 and 0.8815 values between 0.88 and 1.097 values between 1.09 and 1.302 values between 1.30 and 1.515 values between 1.51 and 1.714 values between 1.71 and 1.921 values between 1.92 and 2.131 values between 2.13 and 2.342 values between 2.34 and 2.551 values between 2.76 and 2.971 values between 2.97 and 3.18offnegativepositiveoften

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. ưu
  2. avers
  3. asin
  4. inds
  5. "testing
  6. oke
  7. VEC
  8. idUser
  9. plement
  10. ası

Suppresses

  1. 两年
  2. ties
  3. 几年
  4. 两年前
  5. century
  6. Hier
  7. 的做法
  8. 的高度
  9. ingham
  10. 二十年