Embodied Multi-Modal Agent trained by an LLM from a Parallel TextWorld | Read Paper on Bytez