Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders

Open in new window