[LG] On the Theoretical Understanding of Identifiable Sparse Autoencoders and Beyond
[Peking University & MIT]
https://arxiv.org/abs/2506.15963