Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
4
0
LLMs can unlock advanced spatiotemporal reasoning by dynamically directing a Video World Model to generate crucial visual evidence, outperforming static pipelines.
Images can serve as a powerful standalone medium for reasoning, achieving nearly double the token efficiency of traditional text methods.