外观
论文arXiv 2609.354321 个月前
自进化编码智能体：从数字程序到物理世界智能
原文链接：Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence（arXiv 2609.35432 · 发布日期：2026-09-28）
Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence
提出 Physical Coding 将任务状态与执行表示为代码，构建 HexaAnything，并在 RoboCasa365、PhyBench 和双臂 AgileX 机器人上验证。
核心要点
- VLA/WAM 将观察和指令直映动作，布局或视角小变动易失败且指令泛化差。
- Physical Coding 以代码记录世界状态、关系、约束和进度，并组织规划、验证、恢复与执行。
- HexaAnything 调用感知、规划、控制及 VLA/WAM 工具，依据外部反馈在环决策。
- 在 RoboCasa365 提升 Composite-Unseen 和总体成功率；HexaModel 各划分均优于基线。
验证范围与局限：验证范围为 RoboCasa365、PhyBench，以及一台双臂 AgileX 机器人的物理实验和多数桌面任务。
英文摘要 (Abstract)
Vision-language-action (VLA) and world-action (WAM) models map observations and instructions directly to robot actions. This directness ties a policy to training: minor layout or viewpoint changes cause failure, and instructions generalize poorly. The root cause lies in representation: task requirements, conditions, progress, and failure recovery are implicitly encoded in action sequences, making them difficult to inspect or revise. Digital coding agents offer a precedent: LLMs call tools, verify results, and revise from feedback as executable code. The same working pattern of explicit state, manageable execution, and revisable procedures underlies generalization and long-horizon execution in the physical world, letting physical experience return as reusable programs, memory, or evidence. We propose Physical Coding, representing task state and execution as code. Code as World records objects, relations, constraints, and progress; Code as Policy organizes planning, verification, recovery, and execution. We build HexaAnything, which calls perception, planning, and control tools, including VLA/WAM policies, and makes in-the-loop decisions from external feedback. Verified traces become data and memory, enabling evolution from tools and Harness to model weights, architectures, and ultimately hardware and task design. On RoboCasa365, HexaAnything improves Composite-Unseen and overall success over XR-1 VLA, and its Harness-trained HexaModel beats the base on every split, indicating code traces internalize physical execution. On PhyBench and a dual-arm AgileX robot, the agent autonomously completes physics experiments and most tabletop tasks, often faster than published results. We observe data, model, and tool self-evolution; future work targets weight internalization, autonomous redesign of architectures, languages, representations, and tasks, and deployment in manufacturing and science.
作者：Hongcheng Gao, Jingjing Zhou, Zelin Zheng, Shijia Ge, Jay Zhu, Yazhe Wang, Jianshu Zeng, Xuan Shangguan, Di Wu, Lingyu He
来源
本页事实对应的公开页面，以原文为准。