H2RBench: A Real-to-Sim Benchmark for Evaluating Human-to-Robot Transfer

Chuyang Xiao*, Haotian Zhan*, Sriram Krishna, Peilin Meng, Muhammad Zubair Irshad, Sergey Zakharov, David Held

Comparing human-to-robot (H2R) transfer methods is challenging since existing approaches are evaluated under different task suites, scene layouts, and amounts of robot supervision. H2RBench addresses this with a standardized Real2Sim benchmark built on real human video demonstrations and simulated robot demonstrations across four manipulation tasks reconstructed from real-world scenes. We systematically characterize how representative H2R methods scale with human demonstration data, and show that simulation performance is broadly predictive of real-world performance, with an overall Pearson correlation of r = 0.89, Spearman correlation of ρ = 0.85, and Mean Maximum Rank Violation (MMRV) of 0.06 across method-task configurations.

FreqPolicy: Frequency Autoregressive Visuomotor Policy with Continuous Tokens

Yiming Zhong, Yumeng Liu, Chuyang Xiao, Zemin Yang, Youzhuo Wang, Yufei Zhu, Yujing Sun, Xinge Zhu, Yuexin Ma†

This paper proposes FreqPolicy, a frequency-domain autoregressive visuomotor policy that progressively models hierarchical frequency components with continuous latent representations, achieving superior accuracy and efficiency in robotic manipulation tasks.

DexH2R: A Benchmark for Dynamic Dexterous Grasping in Human-to-Robot Handover

Youzhuo Wang*, Jiayi Ye*, Chuyang Xiao, Yiming Zhong, Heng Tao, Hang Yu, Yumeng Liu, Jingyi Yu, Yuexin Ma†

This paper introduces DexH2R, a real-world dataset for human-to-robot handovers featuring dexterous motions, diverse objects, and rich annotations. Using teleoperation, it captures natural human-like behaviors for robotic learning. We also propose DynamicGrasp, a handover solution, and benchmark state-of-the-art methods, providing key insights for future work.

Optimizing Efficiency of Mixed Traffic through Reinforcement Learning: A Topology-Independent Approach and Benchmark

Chuyang Xiao*, Daiwei Wang*, Xinzheng Tang, Jia Pan, Yuexin Ma†

This paper presents three main contributions to mixed traffic control: (1) the first large-scale real-world benchmark with 444 scenarios from 20 countries; (2) a model-free RL method using Robot Vehicles to optimize traffic flow; (3) superior performance demonstrated in intersections and roundabouts. The benchmark enables realistic policy evaluation in diverse environments.

Carnegie Mellon University

August 2025 - August 2027 (Expected)

Major: Master of Science in Robotics

ShanghaiTech University

September 2021 - June 2025

Major: B.E. in Computer Science

University of California, Berkeley

August 2023 - May 2024

Major: Computer Science (Exchange student in EECS Department)