EmboCoach-Bench: Benchmarking AI Agents on Developing Embodied Robots

TL;DR

This paper introduces EmboCoach-Bench, a benchmark for evaluating autonomous AI agents in developing embodied robotic policies through iterative, feedback-driven workflows, significantly advancing scalable, self-evolving embodied intelligence.

Contribution

It presents a novel benchmark that assesses LLM-based agents' ability to autonomously engineer embodied policies using a dynamic, feedback loop approach across diverse tasks.

Findings

Autonomous agents outperform human baselines by 26.5% success rate.

Environment feedback enhances policy development and reduces performance gaps.

Agents can self-correct and recover from near-failure cases through iterative debugging.

Abstract

The field of Embodied AI is witnessing a rapid evolution toward general-purpose robotic systems, fueled by high-fidelity simulation and large-scale data collection. However, this scaling capability remains severely bottlenecked by a reliance on labor-intensive manual oversight from intricate reward shaping to hyperparameter tuning across heterogeneous backends. Inspired by LLMs' success in software automation and science discovery, we introduce \textsc{EmboCoach-Bench}, a benchmark evaluating the capacity of LLM agents to autonomously engineer embodied policies. Spanning 32 expert-curated RL and IL tasks, our framework posits executable code as the universal interface. We move beyond static generation to assess a dynamic closed-loop workflow, where agents leverage environment feedback to iteratively draft, debug, and optimize solutions, spanning improvements from physics-informed reward design to policy architectures such as diffusion policies. Extensive evaluations yield three critical insights: (1) autonomous agents can qualitatively surpass human-engineered baselines by 26.5\% in average success rate; (2) agentic workflow with environment feedback effectively strengthens policy development and substantially narrows the performance gap between open-source and proprietary models; and (3) agents exhibit self-correction capabilities for pathological engineering cases, successfully resurrecting task performance from near-total failures through iterative simulation-in-the-loop debugging. Ultimately, this work establishes a foundation for self-evolving embodied intelligence, accelerating the paradigm shift from labor-intensive manual tuning to scalable, autonomous engineering in embodied AI field.