Paper 2508.05635
Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation
- Published
- Aug 2025
- Research lab
- Independent
- Citations
- 113
- GitHub
- 568 stars
01 In brief
Summary
Genie Envisioner (GE) is a unified world foundation platform for robotic manipulation, integrating policy learning, evaluation, and simulation within a video-generative framework.
GE-Base, an instruction-conditioned multi-view video diffusion model, is pretrained on ~3,000 hours of data from one million real-world episodes (AgiBot-World-Beta) to capture spatial, temporal, and semantic dynamics.
GE-Act, a lightweight flow-matching decoder, maps latent representations to executable action trajectories, achieving low-latency control (54-step torque in 200ms) and cross-embodiment generalization with only one hour of adaptation data.
GE-Sim, an action-conditioned neural simulator, enables closed-loop policy evaluation and data generation.
EWMBench provides a benchmark suite measuring visual fidelity, physical consistency, and instruction-action alignment.
Experiments show GE-Act outperforms VLA baselines (UniVLA, GR00T N1, π0) on real-world tasks, and GE-Base excels in video generation quality.
The platform is open-sourced.
02 From the paper
Abstract
We introduce Genie Envisioner (GE), a unified world foundation platform for robotic manipulation that integrates policy learning, evaluation, and simulation within a single video-generative framework. At its core, GE-Base is a large-scale, instruction-conditioned video diffusion model that captures the spatial, temporal, and semantic dynamics of real-world robotic interactions in a structured latent space. Built upon this foundation, GE-Act maps latent representations to executable action trajectories through a lightweight, flow-matching decoder, enabling precise and generalizable policy inference across diverse embodiments with minimal supervision. To support scalable evaluation and training, GE-Sim serves as an action-conditioned neural simulator, producing high-fidelity rollouts for closed-loop policy development. The platform is further equipped with EWMBench, a standardized benchmark suite measuring visual fidelity, physical consistency, and instruction-action alignment. Together, these components establish Genie Envisioner as a scalable and practical foundation for instruction-driven, general-purpose embodied intelligence. All code, models, and benchmarks will be released publicly.