RynnVLA-002: A Unified Vision-Language-Action and World Model

TL;DR

RynnVLA-002 presents a unified model integrating vision, language, action, and world prediction to improve environmental understanding and action planning in robots, outperforming previous models in simulation and real-world tasks.

Contribution

It introduces a novel unified framework combining vision-language-action and world modeling for enhanced robotic perception and control.

Findings

Achieves 97.4% success in LIBERO simulation benchmark.

Increases real-world robot success rate by 50%.

Demonstrates mutual enhancement of vision-language-action and world models.

Abstract

We introduce RynnVLA-002, a unified Vision-Language-Action (VLA) and world model. The world model leverages action and visual inputs to predict future image states, learning the underlying physics of the environment to refine action generation. Conversely, the VLA model produces subsequent actions from image observations, enhancing visual understanding and supporting the world model's image generation. The unified framework of RynnVLA-002 enables joint learning of environmental dynamics and action planning. Our experiments show that RynnVLA-002 surpasses individual VLA and world models, demonstrating their mutual enhancement. We evaluate RynnVLA-002 in both simulation and real-world robot tasks. RynnVLA-002 achieves 97.4% success rate on the LIBERO simulation benchmark without pretraining, while in real-world LeRobot experiments, its integrated world model boosts the overall success rate by 50%.