aiqbench analysis
This paper introduces AgiBot World, a massive dataset of over one million robot trajectories designed to train generalist policies for scalable embodied systems. The authors demonstrate that pre-training on this diverse data yields significant performance gains in dexterous and long-horizon tasks compared to existing benchmarks like Open X-Embodiment. By open-sourcing the platform and models, the work aims to standardize high-quality robot data collection and accelerate progress toward general-purpose robotic intelligence.
AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems
arXiv:2503.06669v5 Announce Type: replace-cross Abstract: We explore how scalable robot data can address real-world challenges for generalized robotic manipulation. Introducing AgiBot World, a large-scale platform comprising over 1 million trajectories across 217 tasks in five deployment scenarios, we achieve an order-of-magnitude increase in data scale compared to existing datasets. Accelerated by a standardized collection pipeline with human-in-the-loop verification, AgiBot World guarantees high-quality and diverse data distribution. It is extensible from grippers to dexterous hands and visuo-tactile sensors for fine-grained skill acquisition. Building on top of data, we introduce Genie Operator-1 (GO-1), a novel generalist policy that leverages latent action representations to maximize data utilization, demonstrating predictable performance scaling with increased data volume. Policies pre-trained on our dataset achieve an average performance improvement of 30% over those trained on Open X-Embodiment, both in in-domain and out-of-distribution scenarios. GO-1 exhibits exceptional capability in real-world dexterous and long-horizon tasks, achieving over 60% success rate on complex tasks and outperforming prior RDT approach by 32%. By open-sourcing the dataset, tools, and models, we aim to democratize access to large-scale, high-quality robot data, advancing the pursuit of scalable and general-purpose intelligence.
Abstract:We explore how scalable robot data can address real-world challenges for generalized robotic manipulation. Introducing AgiBot World, a large-scale platform comprising over 1 million trajectories across 217 tasks in five deployment scenarios, we achieve an order-of-magnitude increase in data scale compared to existing datasets. Accelerated by a standardized collection pipeline with human-in-the-loop verification, AgiBot World guarantees high-quality and diverse data distribution. It is extensible from grippers to dexterous hands and visuo-tactile sensors for fine-grained skill acquisition. Building on top of data, we introduce Genie Operator-1 (GO-1), a novel generalist policy that leverages latent action representations to maximize data utilization, demonstrating predictable performance scaling with increased data volume. Policies pre-trained on our dataset achieve an average performance improvement of 30% over those trained on Open X-Embodiment, both in in-domain and out-of-distribution scenarios. GO-1 exhibits exceptional capability in real-world dexterous and long-horizon tasks, achieving over 60% success rate on complex tasks and outperforming prior RDT approach by 32%. By open-sourcing the dataset, tools, and models, we aim to democratize access to large-scale, high-quality robot data, advancing the pursuit of scalable and general-purpose intelligence.
Project website: this https URL. Github repo: this https URL. The author list is ordered alphabetically by surname, with detailed contributions provided in the appendix
Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2503.06669 [cs.RO]
(or arXiv:2503.06669v5 [cs.RO] for this version)
Full text under the publisher's open licence · 225 words
Related coverage
Quantum Computing Bolsters Artificial Intelligence Against Malicious Manipulation
Quantum Zeitgeist · May 30, 2026
Berkeley and Heiserman as an Unexhausted Architecture for Embodied Machine Intelligence
arXiv · Jul 21, 2026
Beyond Dexterity: Why Contact May Define the Next Era of Robotics
IEEE Spectrum · Jun 9, 2026
Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning
arXiv · Jul 21, 2026