Towards Long-horizon Embodied Agents with Tool-Aligned Vision-Language-Action Models

TL;DR

This paper introduces VLAs-as-Tools, a modular approach combining a high-level vision language model with specialized tools for long-horizon embodied tasks, improving planning and execution success.

Contribution

It proposes a novel framework that distributes planning and execution across a VLM agent and specialized VLA tools, with a new interface and training method for tool alignment.

Findings

Improves success rate by 4.8 points on LIBERO-Long.

Enhances success by 23.1 points on RoboTwin.

Increases invocation fidelity by 15 points.

Abstract

Vision-language-action (VLA) models are effective robot action executors, but they remain limited on long-horizon tasks due to the dual burden of extended closed-loop planning and diverse physical operations. We therefore propose VLAs-as-Tools, a strategy that distributes this burden across a high-level vision language model (VLM) agent for temporal reasoning and a family of specialized VLA tools for diverse local physical operations. The VLM handles scene analysis, global planning, and recovery, while each VLA tool executes a bounded subtask. To tightly couple agent planning with VLA tool execution in long-horizon tasks, we introduce a VLA tool-family interface that exposes explicit tool selection and in-execution progress feedback, enabling efficient event-triggered agent replanning without continuous agent polling. To obtain diverse specialized VLA tools that faithfully follow agent invocations, we further propose Tool-Aligned Post-Training (TAPT), which constructs invocation-aligned training units for instruction following and adopts tool-family residual adapters for efficient tool specialization. Experiments show that VLAs-as-Tools improves the success rate of $\pi_{0.5}$ by 4.8 points on LIBERO-Long and 23.1 points on RoboTwin, and further enhances invocation fidelity by 15.0 points as measured by Non-biased Rate. Code will be released.