Neuron Explorer

Qwen3-0.6B

year digits in copyright and dates

A neuron in layer 8 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, especially in copyright lines and dates such as "Copyright 2014" or "Last edited 2020". Turning it up made the model write repeated lines about years and numbers, which fits 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

    may be able to formulate alternatives that may satisfy those interests.⏎The Questioning Process During Bargaining Interactions By Charles B. Craver⏎Copyright © 2014 Charles B. Craver⏎Copyright © 2014 The

  2. #2 peak

    San Antonio vied with each other for trade, commerce and business.⏎George R. Woolfolk⏎Prairie View A&M College⏎Rights© 2021 Prairie View A & M University⏎This work is licensed under a Creative

  3. #3 peak

    Delivery: Yes⏎- Store Delivery: Yes⏎Prices are for internet purchases only. Prices and availability in WHSmith Stores may vary significantly⏎© Copyright 2013 - 2017 WHSmith and its suppliers.⏎WHSmith

  4. #4 peak

    ⏎Publication Type: Peer Reviewed Journal⏎Publication Acceptance Date: February 14, 2008⏎Publication Date: May 16, 2008⏎Citation: Davis, L.M., Spackman

  5. #5 peak

    under general anesthesia in<|endoftext|>BBB - A+ Rating⏎- the best there is⏎An excerpt from www.HouseOfNames.com archives copyright © 2000 - 2014⏎Origins Available: English, Irish

  6. #6 peak

    friendly Cite/link Email Feedback|⏎|Title Annotation:||British Empire|⏎|Article Type:||Brief Article|⏎|Date:||Jan 24, 20

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 most common way of dealing with a problem is to use a mirror, but even if you can, it's still the problem of using a mirror. It's the same as a child's first move. But why? Well, the

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

    ezers or a special tick-removing tool. Grasp the tick as close to the skin as possible. If you don’t have tweezers, protect your fingers with a tissue.⏎- Pull the<|endoftext|>Listen to radar echoes from satellites

  2. #2 peak

    nests, but the Son of Man has no place to rest."⏎American Standard Version⏎And Jesus said unto him, The foxes have holes, and the birds of the heaven have nests; but the Son of man hath not where to lay his

  3. #3 peak

    child and/or the apprehension of a suspect.<|endoftext|>In 1906, Italian economist Vilfredo Pareto made the famous observation that 20% of the population owned 80% of the property in Italy, later

  4. #4 peak

    to use caution at all times. These cats’ stealth and adaptability allow them to survive well in a wide environment.⏎They’ve been known to jump 20 feet in one leap. The puma creeps up on its prey until it

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. L6:932 excites
  2. L4:1291 excites second digit of years 20xx
  3. L1:1675 excites
  4. L1:1862 weakly excites
  5. L0:1529 weakly excites
  6. L5:236 weakly suppresses
  7. L4:973 weakly suppresses
  8. L5:2476 weakly excites
  9. L2:1334 weakly suppresses
  10. L0:329 weakly suppresses the digit 1 in years
  11. L2:2289 weakly suppresses
  12. L4:318 weakly excites the digit 9 in years 1900s
  13. L6:1747 weakly excites technical word fragments, splicing and isotopes
  14. L1:3021 weakly suppresses
  15. L5:881 weakly suppresses
  16. L5:871 weakly suppresses
  17. L5:60 weakly excites
  18. L0:982 weakly excites
  19. L1:1960 weakly excites
  20. L5:1207 weakly suppresses 'of' after abstract nouns
  21. L5:1051 weakly excites
  22. L7:3031 weakly excites
  23. L3:2549 weakly suppresses
  24. L0:2572 weakly excites

Feeds later layers

  1. L18:1551 excites middle digit of years in the 2000s and 2010s
  2. L13:2285 excites
  3. L18:2542 excites
  4. L9:349 excites the spot before a year
  5. L18:1865 excites
  6. L18:2103 excites
  7. L12:1264 excites gap before a year in dates
  8. L10:220 excites hyphen inside compound words
  9. L14:1700 excites
  10. L17:1476 excites
  11. L14:1253 excites
  12. L10:152 excites
  13. L10:456 weakly excites
  14. L13:2971 weakly excites
  15. L16:256 weakly excites number right after 'age of' or 'ages'
  16. L22:1279 weakly excites
  17. L14:395 weakly excites
  18. L10:1784 weakly excites
  19. L9:466 weakly excites
  20. L17:126 weakly excites
  21. L11:300 weakly excites
  22. L12:2691 weakly suppresses
  23. L9:811 weakly excites
  24. L15:368 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.

29 values between -0.87 and -0.49963 values between -0.49 and -0.102764 values between -0.10 and 0.29108 values between 0.29 and 0.689 values between 0.68 and 1.074 values between 1.07 and 1.461 values between 1.46 and 1.844 values between 2.23 and 2.621 values between 3.01 and 3.402 values between 3.40 and 3.783 values between 3.78 and 4.171 values between 4.56 and 4.953 values between 4.95 and 5.343 values between 5.72 and 6.111 values between 6.89 and 7.281 values between 7.66 and 8.051 values between 8.05 and 8.444 values between 8.44 and 8.832 values between 8.83 and 9.221 values between 9.60 and 9.991 values between 11.16 and 11.541 values between 12.71 and 13.101 values between 14.26 and 14.65offnegativepositiveoften

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. Sept
  2. sept
  3. go
  4. ника
  5. lets
  6. /tinyos
  7. ikt
  8. Sept
  9. 机能
  10. yleft

Suppresses

  1. ancode
  2. tres
  3. verständlich
  4. mant
  5. weil
  6. earer
  7. urname
  8. egen
  9. ultipart
  10. .addField