Grounding Sim-to-Real Generalization in Dexterous Manipulation: An Empirical Study with Vision-Language-Action Models

TL;DR

This paper empirically investigates the key factors affecting the transfer of dexterous manipulation policies from simulation to real-world, focusing on Vision-Language-Action models, and provides a standardized evaluation protocol and benchmark.

Contribution

It offers a comprehensive empirical analysis of Sim-to-Real transfer factors for VLA models and introduces a standardized evaluation protocol and benchmark for dexterous manipulation.

Findings

Multi-level domain randomization impacts transfer performance.

Photorealistic rendering improves real-world task success.

Physics-realistic modeling enhances policy robustness.

Abstract

Learning a generalist control policy for dexterous manipulation typically relies on large-scale datasets. Given the high cost of real-world data collection, a practical alternative is to generate synthetic data through simulation. However, the resulting synthetic data often exhibits a significant gap from real-world distributions. While many prior studies have proposed algorithms to bridge the Sim-to-Real discrepancy, there remains a lack of principled research that grounds these methods in real-world manipulation tasks, particularly their performance on generalist policies such as Vision-Language-Action (VLA) models. In this study, we empirically examine the primary determinants of Sim-to-Real generalization across four dimensions: multi-level domain randomization, photorealistic rendering, physics-realistic modeling, and reinforcement learning updates. To support this study, we design a comprehensive evaluation protocol to quantify the real-world performance of manipulation tasks. The protocol accounts for key variations in background, lighting, distractors, object types, and spatial features. Through experiments involving over 10k real-world trials, we derive critical insights into Sim-to-Real transfer. To inform and advance future studies, we release both the robotic platforms and the evaluation protocol for public access to facilitate independent verification, thereby establishing a realistic and standardized benchmark for dexterous manipulation policies.