ERA: Transforming VLMs into Embodied Agents via Embodied Prior Learning and Online Reinforcement Learning

ICLR 2026 Conference Submission15788 Authors

Keywords: Embodied agent, VLM reasoning, reinforcement learning

TL;DR: We present ERA, a two-stage training framework including embodied prior learning and online reinforcement learning, enables 3B VLMs to achieve SOTA on both high-level and low-level embodied tasks.

Abstract:

Recent advances in embodied AI highlight the potential of vision language models (VLMs) as agents capable of perception, reasoning, and interaction in complex environments. However, top-performing systems rely on large-scale models that are costly to deploy, while smaller VLMs lack the necessary knowledge and skills to succeed. To bridge this gap, we present $\textit{Embodied Reasoning Agent (ERA)}$, a two-stage framework that integrates prior knowledge learning and online reinforcement learning (RL).
The first stage, $\textit{Embodied Prior Learning}$, distills foundational knowledge from three types of data: (1) Trajectory-Augmented Priors, which enrich existing trajectory data with structured reasoning generated by stronger models; (2) Environment-Anchored Priors, which provide in-environment knowledge and grounding supervision; and (3) External Knowledge Priors, which transfer general knowledge from out-of-environment datasets.
In the second stage, we develop an online RL pipeline that builds on these priors to further enhance agent performance. To overcome the inherent challenges in agent RL, including long horizons, sparse rewards, and training instability, we introduce three key designs: self-summarization for context management, dense reward shaping, and turn-level policy optimization.
Extensive experiments on both high-level planning (EB-ALFRED) and low-level control (EB-Manipulation) tasks demonstrate that ERA-3B surpasses both prompting-based large models and previous training-based baselines. Specifically, it achieves overall improvements of 8.4\% on EB-ALFRED and 19.4\% on EB-Manipulation over GPT-4o, and exhibits strong generalization to unseen tasks. Detailed Ablation studies further validate the effectiveness of different prior datasets and agent RL designs. Overall, ERA offers a practical path toward scalable embodied intelligence, providing methodological insights for future embodied AI systems.

Primary Area: foundation or frontier models, including LLMs

Submission Number: 15788