Neuron Explorer

Qwen3-0.6B

the spot right before a year's last digits

A neuron in layer 23 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

The neuron fires on the blank spot just before the last digits of a year, such as the gap in 'from 1990 to ___2010'. It is strongly negative on the first digit '2' of years like 2001 or 2020. Turning it up gave unrelated text, and turning it down gave repeated 'zero' words.

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

    in school to increase from 75 to 88 percent. MENA achieved the same increase in 2 decades: from 1990 to 2010.2⏎However, it is important to note that regional

  2. #2 peak

    1994) and at Cadamosto⏎seamount (Heleno et al 1999; Grevemeyer et al. 2010). The rectangle shows the position of Fig. 1b

  3. #3 peak

    1912-2006). Among others that he strongly influenced was the U. S. Federal Reserve Chairman from 1987 to 2006, Alan Greenspan.⏎In 1999 Fre

  4. #4 peak

    isolation of small populations are major concerns in the conservation of large carnivores (Clark et al., 1996; Singleton & Lehmkuhl, 2001). The loss of habitat and connectivity between populations, compounded by

  5. #5 peak

    D member countries. The STEMG was originally formed in 1983 as a working group to develop AECMA Simplified English and subsequently, in 2004, when ASD was created,

  6. #6 peak

    every few days (Salaman and Giles 1995, P. Boesman in litt. 1998, Snyder et al. 2000) on the relatively well-watched San Lorenzo Ridge. It

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:000000000000000000000000000000000000000000000000

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

    change. Even by California standards; it’s ambitious, calling for a 36% reduction in greenhouse gas emissions from 2005 levels by 2020 (statewide emissions decreased 5.3% between 2

  2. #2 peak

    e-learning market in the United States is expected to grow $6.8 billion by 2015, up from $2.9 billion in 2010. According to Sanjeev Ahuja, the vice president of

  3. #3 peak

    all that remains of the giant Buddhas carved into the cliffs are the indentations that once housed them. The Taliban blew these figures up with dynamite in 2001, viewing them as idols from a past they despise

  4. #4 peak

    the issue by modifying the ALA<|endoftext|>Note: I published a previous version of this article on The Change Blog (now called Possibility of Change) in 2011.⏎What is life? Is it the state of physical existence,

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. L21:2497 strongly suppresses
  2. L20:1520 excites
  3. L19:2444 excites
  4. L20:2472 excites
  5. L20:1607 excites
  6. L21:2903 suppresses
  7. L21:741 suppresses
  8. L17:2118 suppresses
  9. L11:2530 excites
  10. L17:1476 excites
  11. L0:305 weakly excites
  12. L20:1355 weakly suppresses
  13. L18:2103 weakly suppresses
  14. L17:1224 weakly excites
  15. L12:697 weakly suppresses
  16. L12:2691 weakly suppresses
  17. L22:906 weakly suppresses
  18. L17:1416 weakly suppresses
  19. L18:1692 weakly suppresses
  20. L13:2285 weakly excites
  21. L20:919 weakly suppresses
  22. L6:1747 weakly excites technical word fragments, splicing and isotopes
  23. L22:504 weakly excites
  24. L19:234 weakly suppresses

Feeds later layers

  1. L25:473 strongly suppresses zeros inside large round numbers
  2. L24:2703 strongly suppresses
  3. L27:153 strongly excites the spot just before a historical number
  4. L26:413 strongly suppresses
  5. L27:862 suppresses
  6. L27:189 suppresses
  7. L27:1153 suppresses
  8. L25:92 suppresses
  9. L25:2960 excites
  10. L25:543 excites
  11. L27:396 suppresses
  12. L27:992 excites
  13. L27:352 excites
  14. L26:834 excites
  15. L27:294 weakly suppresses
  16. L26:682 weakly excites
  17. L27:1202 weakly suppresses
  18. L25:2314 weakly excites the 1 starting a 1900s year
  19. L25:1895 weakly suppresses first digit of a year like 2020
  20. L25:379 weakly excites
  21. L24:969 weakly excites
  22. L24:1318 weakly suppresses
  23. L26:137 weakly excites
  24. L25:2047 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 -247.88 and -240.772 values between -162.56 and -155.452 values between -112.80 and -105.691 values between -105.69 and -98.583 values between -20.38 and -13.275 values between -13.27 and -6.163837 values between -6.16 and 0.9550 values between 0.95 and 8.062 values between 8.06 and 15.173 values between 15.17 and 22.281 values between 22.28 and 29.391 values between 29.39 and 36.50offnegativepositiveoften

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. bidden
  2. pecially
  3. 上世纪
  4. OURS
  5. ieties
  6. ilitation
  7. urally
  8. ictions
  9. емся
  10. ({...

Suppresses

  1. 0
  2. ₀
  3. orio
  4. 0
  5. zero
  6. �
  7. ��
  8. kie
  9. aine
  10. -zero