Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation

2025-08-19

AI Summary

Embodied AI faces a “perception-to-action gap” due to data scarcity and embodiment heterogeneity. Method: This paper introduces a novel embodied reasoning paradigm centered on *pointing* as a unified intermediate representation, formalizing four categories of embodied pointing capabilities. We construct Embodied-Points-200K—the first large-scale embodied pointing dataset—and propose a two-stage reinforcement fine-tuning (RFT) framework, a multi-task reward mechanism, and a cross-modal alignment strategy. Leveraging a 3B-parameter vision-language model, our approach enables end-to-end mapping from spatial understanding to action generation. Results: Our method achieves state-of-the-art performance on 11 embodied pointing tasks; attains 56.2% zero-shot transfer success on SIMPLEREnv; and reaches 87.5% success rate on eight real-world XArm manipulation tasks—outperforming baselines by 62%—while demonstrating strong robustness to visual disturbances.

Technology Category

• Intelligent Robots: Embodied AI
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Generalization in embodied AI is hindered by the "seeing-to-doing gap," which stems from data scarcity and embodiment heterogeneity. To address this, we pioneer "pointing" as a unified, embodiment-agnostic intermediate representation, defining four core embodied pointing abilities that bridge high-level vision-language comprehension with low-level action primitives. We introduce Embodied-R1, a 3B Vision-Language Model (VLM) specifically designed for embodied reasoning and pointing. We use a wide range of embodied and general visual reasoning datasets as sources to construct a large-scale dataset, Embodied-Points-200K, which supports key embodied pointing capabilities. We then train Embodied-R1 using a two-stage Reinforced Fine-tuning (RFT) curriculum with a specialized multi-task reward design. Embodied-R1 achieves state-of-the-art performance on 11 embodied spatial and pointing benchmarks. Critically, it demonstrates robust zero-shot generalization by achieving a 56.2% success rate in the SIMPLEREnv and 87.5% across 8 real-world XArm tasks without any task-specific fine-tuning, representing a 62% improvement over strong baselines. Furthermore, the model exhibits high robustness against diverse visual disturbances. Our work shows that a pointing-centric representation, combined with an RFT training paradigm, offers an effective and generalizable pathway to closing the perception-action gap in robotics.

Problem

• Bridges vision-language comprehension with robotic action primitives
• Addresses data scarcity and embodiment heterogeneity in robotics
• Develops unified pointing representation for generalizable manipulation

Innovation

• Pointing as unified intermediate representation for robotics
• Two-stage reinforced fine-tuning curriculum with multi-task rewards
• 3B vision-language model trained on Embodied-Points-200K dataset