AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models ▲ 7 💬 1 2026-07-04 arXiv HF 原文 GitHub ★3 摘要 （无摘要）