Finding Manifolds With Bilinear Autoencoders
https://arxiv.org/html/2510.16820v1
Ok so they are claiming that sparse autoencoders need input for understanding the representaiotns. THey are proposing bilinear automenders which can study polynomial representations.
Ok what the paper proposes is that we use technique called bilinear autoencoders
© 2026 bsybin