Gemma Scope: helping the safety community shed light...

Towards Monosemanticity: Decomposing Language Models...

Key aspects of Transformer Circuits include: (Plurk ...

哇，有AI幫忙，讀paper真的變得簡單很多。什麼看不懂就問就好。

A Mathematical Framework for Transformer Circuits

還可以用AI幫忙，好有趣！

我今天也打算來看看Gemmascope

psywimac: 哦哦哦，有想到什麼use case嗎？

New LLM Pre-training and Post-training Paradigms

還沒有，目前在往解釋性的方向想就是了