TempR1：通过时序感知多任务强化学习提升多模态大模型的时序理解能力
TempR1: Improving Temporal Understanding of MLLMs via Temporal-Aware Multi-Task Reinforcement Learning
把 GRPO 强化学习扩展到多任务时序训练，并按区间-实例对应关系设计定位奖励，可直接用于长视频时序定位与时间敏感问答的 MLLM 后训练。
原文依据 build upon the Group Relative Policy Optimization (GRPO) algorithm for stable cross-task optimization
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论增强多模态大模型的时序理解能力对长视频分析至关重要，可支撑时序定位与时间敏感问答等任务。尽管已有工作探索了将强化学习用于时序推理，但现有方法往往局限于特定任务或数据集，难以在不同时序场景间泛化。为此，我们提出 TempR1，一种基于强化学习的方法，通过多任务训练增强多模态大模型的时序理解。我们构建了覆盖多种时序结构的多任务语料库，并在 Group Relative Policy Optimization（GRPO）算法基础上实现稳定的跨任务优化。时序任务被划分为三种区间-实例对应类型，并针对每种类型设计了定制化的定位奖励。该设计使 TempR1 能够捕捉细粒度的时序依赖并适应不同的时序模式。实验表明，TempR1 在多个基准上取得最优性能。互补任务间的联合优化进一步产生显著协同效应，同时提升泛化能力与单任务性能，从而为多模态大模型中的时序推理提供了一种可扩展的范式。
英文原文摘要 1234 字符
Enhancing temporal understanding of MLLMs is essential for long-form video analysis, supporting tasks such as temporal localization and time-sensitive question answering. While reinforcement learning (RL) has been explored for temporal reasoning, existing approaches are often limited to specific tasks or datasets, hindering generalization across diverse temporal scenarios. To address this, we present TempR1, an RL-based method that strengthens MLLMs' temporal comprehension through multi-task training. We construct a multi-task corpus covering diverse temporal structures and build upon the Group Relative Policy Optimization (GRPO) algorithm for stable cross-task optimization. Temporal tasks are categorized into three interval-instance correspondence types, with tailored localization rewards designed for each. This design enables TempR1 to capture fine-grained temporal dependencies and adapt to different temporal patterns. Experiments show that TempR1 achieves state-of-the-art performance on multiple benchmarks. Joint optimization across complementary tasks further produces strong synergy, improving both generalization and single-task performance, thereby providing a scalable paradigm for temporal reasoning in MLLMs.