Neuron Explorer

Qwen3-0.6B

weak, mixed: 'on' and 'the' after publishing verbs

A neuron in layer 17 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

Activations on this neuron are small and scattered. The strongest ones fall on words like 'on' after 'appeared' or 'shared', or 'is a' in a definition. It is not clear what ties them together, and turning it up gave text with no obvious link.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    market:⏎The oldest and arguably most successful cryptocurrency on the market is beyond a shadow of a doubt Bitcoin. Developed by the mysterious Satoshi Nakamoto, Bitcoin first appeared on the scene in

  2. #2 peak

    To access this title, visit your library in the app or on the desktop website.⏎Chronicles of the imagination Geographical⏎- Walta Evans "

  3. #3 peak

    locate.<|endoftext|>Science Fair Project Encyclopedia⏎The World Factbook⏎The World Factbook is an annual publication by the Central Intelligence Agency of the United States with basic almanac

  4. #4 peak

    Amazon Web Services⏎|Type of site||Web service, cloud computing|⏎Amazon Web Services (AWS) is a collection of remote computing services (also called web services) that together make up a

  5. #5 peak

    1851.⏎Please post your comment below to share with others. If you'd like to privately share a comment or correction with MMN staff, please use this form.<|endoftext|>Literally the word business means the state of being busy

  6. #6 peak

    2019 after which licensees will begin building commercial models by 2022.⏎If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks<|endoftext|>Potentially harmful traces of

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 number of possible characters in the string in Python, the number of possible characters in the string in the environment of the program's code in the current environment, the number of characters in the string in the current string's string in the string

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

    a project for your teacher, she should have information regarding unit specifications.⏎4Select the toolbars that you will use throughout your drawing. To do this, hover your mouse over an empty space

  2. #2 peak

    and begin to code.⏎Your program should read from the file . The file will describe a single post-cubist squarist

  3. #3 peak

    5 4 3 2 1⏎We need to use the range function to iterate over the array using a for loop. The range() function returns a sequence of numbers starts from a given number to another given number with a given step

  4. #4 peak

    top-rated customer service - once you know, you Newegg.⏎If you are reading this message, Please click this link to reload this page.(Do not use your browser's "Refresh" button). Please email us if you're running the latest

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. L14:2384 excites
  2. L16:2136 excites
  3. L15:1463 excites
  4. L15:1456 suppresses
  5. L12:2117 excites
  6. L15:1464 suppresses
  7. L12:2077 excites
  8. L12:1904 suppresses
  9. L14:1508 suppresses
  10. L15:1914 suppresses
  11. L16:1259 suppresses
  12. L15:1397 excites
  13. L16:1968 excites
  14. L13:1515 excites "your" before doctor or health provider
  15. L15:1514 excites
  16. L16:2984 suppresses
  17. L16:2775 suppresses
  18. L9:2566 excites
  19. L16:958 excites
  20. L8:1639 suppresses
  21. L12:1066 excites
  22. L15:1434 suppresses
  23. L16:1436 excites
  24. L13:851 suppresses

Feeds later layers

  1. L19:1334 suppresses
  2. L18:2040 excites
  3. L19:2401 suppresses
  4. L18:875 suppresses
  5. L21:1206 suppresses
  6. L19:1812 suppresses
  7. L20:716 weakly suppresses
  8. L21:385 weakly suppresses
  9. L18:1801 weakly suppresses
  10. L19:552 weakly suppresses
  11. L19:1344 weakly suppresses
  12. L22:156 weakly suppresses
  13. L18:151 weakly excites
  14. L21:3028 weakly suppresses
  15. L19:2957 weakly suppresses
  16. L20:1615 weakly suppresses
  17. L18:781 weakly suppresses
  18. L18:840 weakly suppresses
  19. L22:827 weakly suppresses
  20. L24:1665 weakly suppresses
  21. L18:2960 weakly excites
  22. L21:1388 weakly suppresses
  23. L27:2633 weakly suppresses
  24. L19:2672 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 -5.15 and -4.961 values between -1.54 and -1.363 values between -1.17 and -0.999 values between -0.81 and -0.6335 values between -0.63 and -0.45181 values between -0.45 and -0.27806 values between -0.27 and -0.091836 values between -0.09 and 0.09809 values between 0.09 and 0.27179 values between 0.27 and 0.4535 values between 0.45 and 0.635 values between 0.63 and 0.814 values between 0.81 and 0.991 values between 1.17 and 1.352 values between 1.53 and 1.711 values between 1.89 and 2.07offnegativepositiveoften

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. losures
  2. amel
  3. alo
  4. →Schema
  5. leys
  6. umo
  7. chein
  8. ements
  9. 〝
  10. alumni

Suppresses

  1. 霉素
  2. _caps
  3. �
  4. orus
  5. browser
  6. INavigation
  7. _oper
  8. naveg
  9. _syntax
  10. (sess