Neuron Explorer

Qwen3-0.6B

the final zero of decades like '1980s'

A neuron in layer 3 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 fires on the last digit '0' of a decade year such as '1880s', '1970s' or '1920s', at the moment the model reads it. Turning it up did not make the text talk about years or decades, so the test did not confirm the pattern.

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

    Pituophis melanoleucus mugitus), eastern coachwh<|endoftext|>Fair Use of Copyrighted Works by Teachers and Students⏎Since the mid-1880s, teachers and researchers have relied on an exception to copyright protection called fair use

  2. #2 peak

    rage was caught up in the newspaper business, because the press was one of the only places where women could voice their opinions. In the late 1890s and into the early 1900s, it was not considered

  3. #3 peak

    the same encryption key for every transaction hasn't been an acceptable practice since the Caesar Cipher was popular in ancient Rome. Furthermore, since the mid-1970s, the DES algorithm was known to be vulnerable to brute-force attacks because of

  4. #4 peak

    The US Congress has had oversight responsibilities over the CIA since the Agency was established in 1947. However, prior to the mid-1970s, oversight responsibilities resided in the Armed Services Committees of both chambers and were

  5. #5 peak

    city of Hollywood, Florida you’ll find a small community, populated mainly by African Americans, known by the name of Liberia.⏎In the early 1920s, a vast amount of land in this area was purchased by a gentleman named

  6. #6 peak

    Caroline had finished her studies in Sweden in the early 1880's, before moving South, to Paris, and settling in the village of Gre

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 am working on a essay about a event, a person, or a idea. The essay should be in the present tense. **Context:** The event of the first century, the story is set in a ancient city, the city

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

    heat to the surrounding rocks. The magma fills and widens cracks and fractures as it rises. The end result is a vertical wall, or dyke,<|endoftext|>-Why is balance important in an ecosystem?⏎-What is the relationship between living and

  2. #2 peak

    only one example of clever tricks used by cephalopods. Check out our full-length feature to find out more about these animals' astounding mental skills.<|endoftext|>- Cooking Time: 0⏎- Servings: 0⏎- Here are

  3. #3 peak

    43],. In the Sogdian sources (early 8th – 9th

  4. #4 peak

    percent of its normal value, a depression that only decayed once the pole's location stabilized.<|endoftext|>- Project plans⏎- Project activities⏎- Legislation and standards⏎- Industry context⏎

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. L2:404 strongly suppresses
  2. L1:217 excites
  3. L0:50 weakly suppresses
  4. L0:132 weakly excites
  5. L1:2342 weakly excites
  6. L0:3000 weakly suppresses
  7. L0:1792 weakly excites
  8. L2:422 weakly suppresses
  9. L2:1091 weakly excites
  10. L0:1296 weakly excites
  11. L1:1017 weakly suppresses
  12. L1:1016 weakly suppresses
  13. L2:228 weakly excites
  14. L1:1343 weakly suppresses
  15. L1:331 weakly suppresses
  16. L0:3065 weakly excites
  17. L1:2258 weakly suppresses
  18. L2:75 weakly suppresses
  19. L2:1727 weakly suppresses
  20. L2:152 weakly suppresses
  21. L1:2200 weakly suppresses
  22. L1:937 weakly excites
  23. L1:1062 weakly suppresses
  24. L0:2518 weakly suppresses

Feeds later layers

  1. L23:393 strongly excites
  2. L25:577 strongly suppresses
  3. L20:344 strongly suppresses hyphen inside compound words
  4. L8:3004 strongly excites
  5. L4:490 strongly excites decade digit in years like 1930s
  6. L20:1513 strongly suppresses
  7. L18:1997 suppresses
  8. L27:2408 excites
  9. L20:961 excites
  10. L20:1385 suppresses
  11. L22:265 suppresses
  12. L26:1008 excites
  13. L25:1757 suppresses
  14. L5:97 excites
  15. L23:474 suppresses
  16. L25:231 excites
  17. L21:1145 suppresses
  18. L4:384 suppresses
  19. L24:318 excites
  20. L25:535 suppresses commas inside number lists
  21. L24:248 suppresses
  22. L21:690 suppresses
  23. L22:512 suppresses
  24. L18:853 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 -0.74 and -0.682 values between -0.63 and -0.582 values between -0.58 and -0.522 values between -0.52 and -0.475 values between -0.47 and -0.4219 values between -0.42 and -0.3623 values between -0.36 and -0.3137 values between -0.31 and -0.26106 values between -0.26 and -0.20190 values between -0.20 and -0.15365 values between -0.15 and -0.10602 values between -0.10 and -0.04971 values between -0.04 and 0.01712 values between 0.01 and 0.06451 values between 0.06 and 0.12211 values between 0.12 and 0.17105 values between 0.17 and 0.2251 values between 0.22 and 0.2819 values between 0.28 and 0.3314 values between 0.33 and 0.385 values between 0.38 and 0.443 values between 0.44 and 0.493 values between 0.49 and 0.541 values between 0.54 and 0.602 values between 0.60 and 0.651 values between 0.65 and 0.701 values between 0.92 and 0.971 values between 1.03 and 1.081 values between 1.08 and 1.131 values between 1.24 and 1.291 values between 1.35 and 1.40offnegativepositiveoften

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. s
  2. ات
  3. es
  4. sut
  5. y
  6. اتهم
  7. sen
  8. sampling
  9. yı
  10. sar

Suppresses

  1. naire
  2. noch
  3. -contrib
  4. naires
  5. /id
  6. .shtml
  7. 缪
  8. -abs
  9. gắng
  10. Toll