Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence

2026-09-28

AI Summary

This study addresses the limitations of Vision-Language-Action (VLA) models—namely poor generalization, difficulty in correction, and failure in long-horizon tasks—arising from implicit encoding. To overcome these challenges, this work proposes a Physical Coding paradigm that explicitly encodes physical states and policies into executable programs. Methodologically, we construct the HexaAnything agent to bridge code generation with the physical world by integrating large language model-based code synthesis with vision-language-action models. A Harness training mechanism is further introduced to support tool invocation and feedback-driven decision-making, establishing a self-evolving closed loop among data, models, and tools. Experimental results demonstrate that the proposed approach significantly outperforms the XR-1 baseline and existing state-of-the-art methods on the RoboCasa365 benchmark, while successfully executing complex physical experiments autonomously.

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.

Problem

• Vision-Language-Action models
• Physical-world intelligence
• Representation learning
• Generalization
• Long-horizon execution

Innovation

• Physical Coding
• Self-Evolving Agents
• Vision-Language-Action
• Code as Policy
• Robotic Manipulation