1 paper
Theo W. Costain, Kejie Li, Victor A. Prisacariu
Prior works have demonstrated that implicit representations trained only for reconstruction tasks typically generate encodings that are not useful for semantic tasks. In this work,…