Ground, Cover, and Refine: Evidence-Centric Frame Selection for Long-Video Question Answering

作者：Fan Wei, Siru Zhong, Runmin Dong, Miao Yang, Zhaoyang Luo, Haohuan Fu · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-08-04