The Year in AI Papers
The year/Qwen

Paper 2605.30280

Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments

Qiuyue Wang, Mingsheng Li, Jian Guan, Jinhui Ye, Sicheng Xie, Yitao Liu, Junhao Chen, Zhixuan Liang, Jie Zhang, Xintong Hu, Xuhong Huang, Pei Lin, et al.

Published
May 2026
Research lab
Qwen
Citations
20
GitHub
Not linked

01 In brief

Summary

Qwen-VLA is a unified embodied foundation model that extends the Qwen3.5-4B vision-language backbone with a DiT-based flow-matching action decoder, enabling a single model to handle manipulation, navigation, and trajectory prediction across diverse tasks, environments, and robot embodiments.

It uses embodiment-aware prompt conditioning and a unified action-and-trajectory representation to support multiple platforms without per-embodiment policies.

Training follows a four-stage recipe: text-to-action DiT pretraining (T2A), continued pretraining (CPT), supervised fine-tuning (SFT), and reinforcement learning (RL).

The pretraining mixture includes robot manipulation trajectories (74.2%), egocentric human data, synthetic simulation data, navigation data, and auxiliary vision-language data.

Qwen-VLA-Instruct achieves state-of-the-art results: 97.9% on LIBERO, 73.7% on Simpler-WidowX, 86.1/87.2% on RoboTwin-Easy/Hard, 69.0% OSR on R2R, and 59.6% SR on RxR.

In real-world ALOHA experiments, it attains 76.9% average OOD success, and 26.6% zero-shot success on DOMINO dynamic manipulation.

Ablations show T2A improves SFT success by +10.2 pp, and RL provides further gains, especially in the training environment, with mild positive transfer to other benchmarks.

The model demonstrates strong out-of-distribution generalization across scene, object, lighting, and embodiment variations, supporting the view that heterogeneous embodied tasks share a common action-and-trajectory prediction structure.

02 From the paper

Abstract

Embodied intelligence is often studied through specialized models for individual tasks such as manipulation or navigation, resulting in fragmented capabilities and limited generalization across tasks, environments, and robot embodiments. In this work, we study whether heterogeneous embodied decision-making problems can be unified within a single vision-language-action model. We present Qwen-VLA, a unified embodied foundation model that extends Qwen's vision-language modeling stack from perception, understanding, and reasoning to continuous action and trajectory generation through a DiT-based action decoder. Qwen-VLA is trained with a large-scale joint pretraining recipe over diverse data sources, including robotics manipulation trajectories, human egocentric demonstrations, synthetic simulation data, vision-and-language navigation data, trajectory-centric supervision, and auxiliary vision-language data. To support multiple robot platforms, we introduce embodiment-aware prompt conditioning, where robot-specific textual descriptions specify the current embodiment and control convention. We further cast manipulation, navigation, and trajectory prediction into a unified action-and-trajectory prediction framework, enabling transferable visual grounding, spatial reasoning, and continuous action generation across robot morphologies, task families, and environments. Experiments on manipulation, navigation, and trajectory-centric benchmarks show consistent multi-task performance and out-of-distribution generalization under variations in scene layout, background, lighting, object configuration, and robot embodiment. Qwen-VLA-Instruct achieves 97.9% on LIBERO, 73.7% on Simpler-WidowX, 86.1%/87.2% on RoboTwin-Easy/Hard, 69.0% OSR on R2R, 59.6% SR on RxR, 76.9% average OOD success in real-world ALOHA experiments, and 26.6% zero-shot success on DOMINO dynamic manipulation.