paper/arxiv/2608.27033

Robotics Dataset

Riemann-1.0 is a causal autoregressive World Action Model that unifies robot policy execution and world simulation, trained on 200K+ hours of heterogeneous embodied data including egocentric human videos, handheld-gripper demonstrations, and robot trajectories, achieving state-of-the-art performance on simulation and real-world manipulation benchmarks.

Catalog metadata

Repository owner

Haofeng Sun, Jiangbo Pei, Fei Kang et al.

Source

Research paper

Task

Riemann-1.0: An Embodied World Action Model for Physical AI

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.