Search papers, labs, and topics across Lattice.
1
0
3
Video-LLMs can hallucinate and perform *worse* with chain-of-thought reasoning due to "visual anchor drifting," but a simple frame repetition strategy guided by a learned scoring function can fix it.
