arXiv 2604.281922026-05-07cs.ROcs.CV

LaST-R1：通过自适应物理潜在推理强化机器人操作

LaST-R1: Reinforcing Robotic Manipulation via Adaptive Physical Latent Reasoning

The Chinese University of Hong Kong（香港中文大学）；State Key Laboratory of Multimedia Information Processing, School of Computer Science, Peking University（北京大学多媒体信息处理国家重点实验室，计算机学院）；Simplexity Robotics Project（Simplexity机器人项目）

Hao Chen, Jiaming Liu, Zhonghao Yan, Nuowei Han, Renrui Zhang, Chenyang Gu, Jialin Gao, Ziyu Guo, Siyuan Qian, Yinxi Wang, Peng Jia, Shanghang Zhang, Pheng-Ann Heng

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出LaST-R1，一种基于强化学习的后训练框架，通过联合优化潜在推理过程和动作生成，提升机器人在动态环境中的适应性和泛化能力。

摘要：

机器人基础模型需要对复杂视觉场景进行推理，以在动态环境中执行自适应动作。尽管最近关于潜在推理视觉-语言-动作（VLA）模型的研究展示了捕捉细粒度物理动态的能力，但它们仍主要局限于静态模仿学习，严重限制了其适应性和泛化能力。本文提出LaST-R1，一种新颖的强化学习（RL）后训练框架，旨在有效利用“潜在推理前行动”策略。具体来说，我们提出Latent-to-Action Policy Optimization（LAPO），一种核心RL算法，联合优化潜在推理过程和动作生成。通过显式嵌入潜在链式思维（CoT）推理到RL优化循环中，LAPO激发了深刻的物理世界建模，从而驱动在交互环境中的稳健执行。此外，引入了一种自适应潜在Co T机制，使策略能够根据不同的环境状态动态调节其推理范围。实验表明，LaST-R1在LIBERO基准测试中仅需一次单次监督预热即可达到99.9%的平均成功率，显著提高了收敛速度和性能，优于先前最先进的（SOTA）方法。在现实世界部署中，LaST-R1在四个复杂任务中，包括单臂和双臂设置，平均比SOTA监督微调方法提高了22.5%。最后，LaST-R1在模拟和现实环境中的泛化能力均表现出色。

英文摘要：

Robotic foundation models require reasoning over complex visual scenes to execute adaptive actions in dynamic environments. While recent studies on latent-reasoning Vision-Language-Action (VLA) models have demonstrated the capability to capture fine-grained physical dynamics, they remain predominantly confined to static imitation learning, severely limiting their adaptability and generalization. In this paper, we present LaST-R1, a novel reinforcement learning (RL) post-training framework designed to effectively harness "latent reasoning-before-acting" policies. Specifically, we propose Latent-to-Action Policy Optimization (LAPO), a core RL algorithm that jointly optimizes the latent reasoning process and the action generation. By explicitly embedding latent Chain-of-Thought (CoT) reasoning directly within the RL optimization loop, LAPO stimulates profound physical world modeling, which in turn drives robust execution in interactive environments. Furthermore, an adaptive latent CoT mechanism is introduced, allowing the policy to dynamically modulate its reasoning horizon based on diverse environment states. Experiments show that LaST-R1 achieves a near-perfect 99.9% average success rate on the LIBERO benchmark with only one-shot supervised warm-up, significantly improving convergence speed and performance over prior state-of-the-art (SOTA) methods. In real-world deployments, LaST-R1 yields up to a 22.5% average improvement over SOTA supervised fine-tuning approach across four complex tasks, including both single-arm and dual-arm settings. Finally, LaST-R1 demonstrates strong generalization across simulated and real-world environments.

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。