Persistent Object Narratives for Token-Efficient Video Language Models

作者：Junzhe Chen, Siyuan Meng, Xiaojie Guo · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-08-06