DexVerse: A Modular Benchmark for Multi-Task, Multi-Embodiment Dexterous Manipulation

2026-07-09

AI Summary

Existing dexterous manipulation benchmarks are limited in task diversity, robot embodiment coverage, and visual controllability, hindering the evaluation of cross-task and cross-embodiment generalization. This work proposes the first large-scale, modular dexterous manipulation benchmark, systematically integrating 100 tasks, three robotic arms, and six dexterous hands, with support for vision-conditioned configurations and extensibility to new tasks. The benchmark features modular task design, configurable rendering, and a VR teleoperation interface, providing synchronized multimodal data including RGB, depth, point clouds, proprioception, and state observations. Using this benchmark, we evaluate representative methods—Diffusion Policy, DP3, and OpenVLA—on 19 tasks, revealing critical challenges in current models regarding task generalization and visual robustness.

Abstract

Building general-purpose dexterous manipulation policies requires benchmarks that go beyond isolated tasks to systematically evaluate policies across diverse interaction modes, sensory conditions, and robot embodiments. However, existing benchmarks remain limited in task and data diversity, embodiment coverage, or controllable visual variation, hindering studies of cross-task and cross-embodiment generalization. We present DexVerse, a large-scale and modular benchmark for dexterous manipulation. DexVerse includes 100 tasks spanning a broad range of manipulation skills, including object grasping and relocation, articulated-object interaction, functional tool use, bimanual coordination, non-prehensile control, contact-rich behaviors, multi-goal execution, and long-horizon multi-stage task completion. It supports 3 robot arms and 6 dexterous hands, and is extensible to new tasks, assets, and embodiments. To evaluate visuomotor generalization, DexVerse provides configurable visual variations in textures, background, lighting, and camera viewpoints. We further provide a VR-based teleoperation interface and 3,180 demonstrations with synchronized proprioceptive, RGB, depth, point-cloud, and state observations. We benchmark representative methods, including Diffusion Policy, DP3, OpenVLA, and $π_{0.5}$, across 19 tasks. Results reveal substantial challenges in task generalization and visuomotor robustness, establishing DexVerse as a promising testbed for general-purpose dexterous manipulation. Project page: https://ycyao216.github.io/DexVerse.site

Problem

• dexterous manipulation
• benchmark
• multi-task
• multi-embodiment
• generalization

Innovation

• dexterous manipulation
• multi-embodiment
• visuomotor generalization
• modular benchmark
• teleoperation demonstrations