Search papers, labs, and topics across Lattice.
4
0
6
17
RaDiCal reveals that relying solely on attention saliency can misguide token pruning, leading to a 39-45% reduction in FLOPs while enhancing ranking performance across key datasets.
Image corruptions can drastically impair retrieval performance, revealing that clean image-only queries often outperform joint retrieval in snap-and-ask scenarios.
EviRank redefines multimodal image re-ranking by transforming complex queries into structured evidence, achieving unprecedented accuracy and efficiency.
An 8B model can rival Gemini 3-Pro in generating detailed, temporally-aware scripts from long-form video, proving that targeted training trumps brute force scaling for narrative comprehension.
