AI 的“读心术”：我们如何才能信任一个“聪明的大脑”？
[LG] Taming Polysemanticity in LLMs: Provable Feature Recovery via Sparse Autoencoders
[Yale University & Shanghai Jiao Tong University]
https://arxiv.org/abs/2506.14002
在小宇宙查看该单集文稿
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.