GenRL: Multimodal-foundation world models for generalization in embodied agents | Read Paper on Bytez