Refer-Agent：面向指代视频目标分割的推理与反思协同多智能体系统
Refer-Agent: A Collaborative Multi-Agent System with Reasoning and Reflection for Referring Video Object Segmentation
把 RVOS 拆成由 MLLM 驱动的多智能体逐步推理流程，用 Questioner-Responder 反思链校验中间结果并回灌下一轮推理，配合由粗到细的帧选择与动态视觉焦点布局；这套无需微调即可接入新 MLLM 的推理框架，可直接迁移到视频问答、时序定位等需要跨帧证据聚合的视觉-语言任务。
原文依据 we propose Refer-Agent, a collaborative multi-agent system with alternating reasoning-reflection mechanisms
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论指代视频目标分割（RVOS）旨在根据文本查询对视频中的目标进行分割。当前方法主要依赖多模态大语言模型（MLLM）的大规模监督微调（SFT）。然而，该范式数据依赖性强，且难以应对 MLLM 的快速演进，可扩展性受限。尽管近期的零样本方法提供了灵活的替代方案，但由于工作流设计过于直接，其性能仍显著落后于基于 SFT 的方法。为解决上述局限，我们提出 Refer-Agent，一个具备交替推理-反思机制的协同多智能体系统。该系统将 RVOS 分解为逐步推理过程。在推理阶段，我们引入由粗到细的帧选择策略，以保证帧的多样性与文本相关性，并提出动态焦点布局以自适应调整智能体的视觉关注区域。此外，我们提出反思链机制，采用提问者-回答者配对生成自反思链，使系统能够校验中间结果并生成反馈，用于下一轮推理的改进。在五个具有挑战性的基准上的大量实验表明，Refer-Agent 显著优于最优方法，包括基于 SFT 的模型与零样本方法。此外，Refer-Agent 具有灵活性，可在不产生任何额外微调成本的情况下快速集成新的 MLLM。代码将开源。
英文原文摘要 1510 字符
Referring Video Object Segmentation (RVOS) aims to segment objects in videos based on textual queries. Current methods mainly rely on large-scale supervised fine-tuning (SFT) of Multi-modal Large Language Models (MLLMs). However, this paradigm suffers from heavy data dependence and limited scalability against the rapid evolution of MLLMs. Although recent zero-shot approaches offer a flexible alternative, their performance remains significantly behind SFT-based methods, due to the straightforward workflow designs. To address these limitations, we propose Refer-Agent, a collaborative multi-agent system with alternating reasoning-reflection mechanisms. This system decomposes RVOS into step-by-step reasoning process. During reasoning, we introduce a Coarse-to-Fine frame selection strategy to ensure the frame diversity and textual relevance, along with a Dynamic Focus Layout that adaptively adjusts the agent's visual focus.Furthermore, we propose a Chain-of-Reflection mechanism, which employs a Questioner-Responder pair to generate a self-reflection chain, enabling the system to verify intermediate results and generates feedback for next-round reasoning refinement. Extensive experiments on five challenging benchmarks demonstrate that Refer-Agent significantly outperforms state-of-the-art methods, including both SFT-based models and zero-shot approaches.Moreover, Refer-Agent is flexible and enables fast integration of new MLLMs without any additional fine-tuning costs. Code will be released.