GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Data

TL;DR

GraspVLA is a foundation model trained on a billion-scale synthetic dataset for robotic grasping, demonstrating strong zero-shot and few-shot generalization capabilities across real and simulated environments.

Contribution

This work introduces a novel synthetic dataset and a unified VLA model pre-trained on it, enabling effective grasping with reduced reliance on real-world data.

Findings

Achieves state-of-the-art zero-shot grasping performance

Demonstrates effective transfer to real-world grasping tasks

Shows strong few-shot adaptability to human preferences

Abstract

Embodied foundation models are gaining increasing attention for their zero-shot generalization, scalability, and adaptability to new tasks through few-shot post-training. However, existing models rely heavily on real-world data, which is costly and labor-intensive to collect. Synthetic data offers a cost-effective alternative, yet its potential remains largely underexplored. To bridge this gap, we explore the feasibility of training Vision-Language-Action models entirely with large-scale synthetic action data. We curate SynGrasp-1B, a billion-frame robotic grasping dataset generated in simulation with photorealistic rendering and extensive domain randomization. Building on this, we present GraspVLA, a VLA model pretrained on large-scale synthetic action data as a foundational model for grasping tasks. GraspVLA integrates autoregressive perception tasks and flow-matching-based action generation into a unified Chain-of-Thought process, enabling joint training on synthetic action data and Internet semantics data. This design helps mitigate sim-to-real gaps and facilitates the transfer of learned actions to a broader range of Internet-covered objects, achieving open-vocabulary generalization in grasping. Extensive evaluations across real-world and simulation benchmarks demonstrate GraspVLA's advanced zero-shot generalizability and few-shot adaptability to specific human preferences. We will release SynGrasp-1B dataset and pre-trained weights to benefit the community.