Multi-Task Learning of Generalizable Representations for Video Action Recognition | Read Paper on Bytez