Search papers, labs, and topics across Lattice.
2
0
4
5
Current MLLMs falter in following complex video instructions, revealing a significant oversight in their evaluation metrics.
FIRM-Video reveals that a checklist-driven approach can significantly enhance the reliability of text-to-video reward modeling, achieving state-of-the-art performance in evaluation metrics.
