EMHO: EMbodied Agent Harness Optimization via Experience Traces

Key points

• Improving embodied agents often focuses on optimizing the underlying model through training, while the surrounding agent harness that controls planning, context, and tool use is typically engineered.1
• We ask whether this harness can instead improve itself directly from experience traces under sparse environmental feedback.1
• To support multiple subtasks with a single harness, we introduce EMHO-Merge, which addresses trade-offs in jointly optimizing a single shared harness across subtasks by using episode-level gains and losses to guide evidence-supported refinement of when and how revised behaviors are applied.1
• We evaluate EMHO on EmbodiedBench across navigation and manipulation tasks, and EMHO consistently improves task success for both Qwen 9B and 27B models.1

Sources (1)

• [1]EMHO: EMbodied Agent Harness Optimization via Experience Traces
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 02:30 PM

  We propose EMbodied Agent Harness Optimization (EMHO), a self-evolving framework that keeps the embodied model frozen and iteratively revises its harness by analyzing execution trajectories and prior harness history.

  Improving embodied agents often focuses on optimizing the underlying model through training, while the surrounding agent harness that controls planning, context, and tool use is typically engineered.

Extractive summary: sentences quoted from the sources.