arXiv 2508.062062026-05-20cs.ROcs.CV

Affordance-R1: 为多模态大语言模型中的通用化 affordance 推理设计的强化学习

Affordance-R1: Reinforcement Learning for Generalizable Affordance Reasoning in Multimodal Large Language Model

The Hong Kong University of Science and Technology (GZ)（香港科技大学（广州））；National University of Singapore（新加坡国立大学）；ShanghaiTech University（上海科技大学）；East China Normal University（华东师范大学）；Nanjing University of Information Science & Technology（南京信息工程大学）；Zhejiang University（浙江大学）；Institute of Automation, Chinese Academy of Science（中国科学院自动化研究所）；Shanghai AI Laboratory（上海人工智能实验室）

Hanqing Wang, Shaoyang Wang, Yiming Zhong, Zemin Yang, Jiamin Wang, Zhiqing Cui, Jiahao Yuan, Yifan Han, Mingyu Liu, Yuexin Ma

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出 Affordance-R1，一种结合认知 CoT 引导的 Group Relative Policy Optimization (GRPO) 的统一 affordance 地标框架，通过强化学习实现零样本泛化和测试时推理能力。

摘要：

Affordance grounding 旨在预测与机器人执行动作相关的物体特定区域。它在人机交互、人-物交互、具身操作和具身感知领域中起着至关重要的作用。现有模型由于缺乏链式思维（CoT）推理能力，往往忽视不同物体间的 affordance 共享，限制了其域外（OOD）泛化和显式推理能力。为了解决这些挑战，我们提出了 Affordance-R1，这是首个集成认知 CoT 引导的 Group Relative Policy Optimization（GRPO）的统一 affordance 地标框架。具体而言，我们设计了一个复杂的 affordance 函数，包含格式、感知和认知奖励，以有效引导优化方向。此外，我们构建了一个高质量的 affordance 中心推理数据集 ReasonAff，以支持训练。通过仅使用强化学习与 GRPO 进行训练，而不使用显式推理数据，Affordance-R1 实现了稳健的零样本泛化，并表现出涌现的测试时推理能力。全面的实验表明，我们的模型优于已建立的方法，并展示了开放世界泛化能力。据我们所知，Affordance-R1 是首个将基于 GRPO 的 RL 与推理结合到 affordance 推理中的方法。我们的方法和数据集已发布在 https://github.com/hq-King/Affordance-R1。

英文摘要：

Affordance grounding focuses on predicting the specific regions of objects that are associated with the actions to be performed by robots. It plays a vital role in the fields of human-robot interaction, human-object interaction, embodied manipulation, and embodied perception. Existing models often neglect the affordance shared among different objects because they lack the Chain-of-Thought(CoT) reasoning abilities, limiting their out-of-domain (OOD) generalization and explicit reasoning capabilities. To address these challenges, we propose Affordance-R1, the first unified affordance grounding framework that integrates cognitive CoT guided Group Relative Policy Optimization (GRPO) within a reinforcement learning paradigm. Specifically, we designed a sophisticated affordance function, which contains format, perception, and cognition rewards to effectively guide optimization directions. Furthermore, we constructed a high-quality affordance-centric reasoning dataset, ReasonAff, to support training. Trained exclusively via reinforcement learning with GRPO and without explicit reasoning data, Affordance-R1 achieves robust zero-shot generalization and exhibits emergent test-time reasoning capabilities. Comprehensive experiments demonstrate that our model outperforms well-established methods and exhibits open-world generalization. To the best of our knowledge, Affordance-R1 is the first to integrate GRPO-based RL with reasoning into affordance reasoning. The code of our method and our dataset is released on https://github.com/hq-King/Affordance-R1.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。