学习拒答：面向视频时序定位中难负相关查询的拒答感知强化微调
Learning to Refuse: Refusal-Aware Reinforcement Fine-Tuning for Hard-Irrelevant Queries in Video Temporal Grounding
把 GRPO 强化微调引入 VTG，用格式、拒答-IoU、解释、查询纠正四类奖励让 LVLM 学会对语义相近但实际无关的难负查询输出拒答，可直接迁移到任何需要「无目标即拒答」的定位/检索式视觉任务。
原文依据 we propose Refusal-Aware Reinforcement Fine-Tuning (RA-RFT) to effectively refuse hard-irrelevant queries in VTG
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频时序定位（VTG）旨在定位视频中与自然语言查询相对应的时序片段。然而，现有 VTG 模型假设相关片段总是存在，导致即使查询与视频无关时也总会预测出一个目标片段。尽管近期方法尝试处理无关查询，但它们只能拒绝与视频完全无关的查询，仍无法处理语义相似但实际并不相关的难负相关查询。为解决这一问题，我们提出拒答感知强化微调（RA-RFT），以在 VTG 中有效拒答难负相关查询。该方法基于组相对策略优化（GRPO）框架，并整合格式、拒答-IoU、解释与查询纠正四项奖励目标，以同时提升相关性判别能力与细粒度语义推理能力。此外，为有效支撑 RA-RFT，我们构建了难负相关 VTG（HI-VTG）数据集，其中包含难负相关查询及其拒答答案。我们在多种相关性感知 VTG 场景中验证了该方法的有效性，包括难负相关 VTG、简单打乱的 RA-VTG 以及人工标注的 RA-VTG 设置。我们还通过将该方法应用于多种基于 LVLM 的 VTG 模型，证明了其可扩展性。
英文原文摘要 1348 字符
Video Temporal Grounding (VTG) aims to localize a temporal segment in a video corresponding to a natural language query. However, existing VTG models assume that a relevant segment always exists, causing them to always predict a target segment even when the query is irrelevant to the video. While recent approaches attempt to handle irrelevant queries, they can only reject those that are entirely unrelated to the video and still fail to handle hard-irrelevant queries that are semantically similar but not actually relevant.To address this, we propose Refusal-Aware Reinforcement Fine-Tuning (RA-RFT) to effectively refuse hard-irrelevant queries in VTG.Our method is based on the Group Relative Policy Optimization (GRPO) framework and integrates four reward objectives--format, refuse-IoU, explain, and query correction--to improve both relevance discrimination and fine-grained semantic reasoning.In addition, to effectively support RA-RFT, we construct a Hard-Irrelevant VTG (HI-VTG) dataset, which includes hard-irrelevant queries and their refusal answers.We demonstrate the effectiveness of our method across various relevance-aware VTG scenarios, including hard-irrelevant VTG, simply-shuffled RA-VTG, and human-annotated RA-VTG settings. We also show that the proposed method is scalable by applying it to various LVLM-based VTG models.