【SHANGHAI, China, June 3, 2026】AGIBOT today announced the release of AGIBOT WORLD 2026 Theme 2: Rich Interaction, a new theme of its real-world embodied AI dataset initiative designed to capture the complex, contact-rich interactions between robots and the physical world.

While many traditional robotics datasets focus primarily on clean, successful task demonstrations, real-world intelligence requires learning from a much broader spectrum of physical experience, including unexpected outcomes, imperfect executions, and edge-case behaviors. AGIBOT WORLD 2026 Theme 2 takes a step beyond “perfect demonstrations” by systematically capturing how robots interact with diverse objects, materials, structures, and environments under realistic physical conditions.

Built entirely from 100% real-world data, the Rich Interaction dataset introduces an exploratory teleoperation paradigm in which operators intentionally guide robots to interact with a wide range of objects across different materials, geometries, mechanical structures, and functional properties. The dataset records not only successful interactions, but also imperfect and highly informative physical events, including missed grasps, collisions, object drops, unstable contacts, and liquid splashes.

This approach is designed to reflect the full distribution of physical interaction in the real world. By capturing both expected and unexpected outcomes, the dataset provides high-fidelity information on contact dynamics, material deformation, object response, and multi-modal feedback. This is data that is often difficult to obtain through simulation or conventional task-oriented demonstrations.

AGIBOT WORLD 2026 Theme 2 is specifically developed to address a critical data gap in the embodied AI community. ‘Rich interaction’dataset is supposed to support research in world models, neural simulators, physics-informed perception, and robust representation learning. By exposing models to the nuanced and sometimes chaotic nature of real-world physics, the dataset aims to help next-generation embodied AI systems better understand, predict, and respond to the physical world.

The launch of Theme 2 follows the release of AGIBOT WORLD 2026 Theme 1, which focused on imitation learning data collected from real-world robot operations. Together, the two phases reflect AGIBOT’s broader commitment to building an open, high-quality data foundation for embodied AI, supporting researchers and developers as they explore more scalable paths toward physical-world intelligence.

Looking ahead, AGIBOT will continue to expand AGIBOT WORLD 2026 with additional data themes, benchmarks, and research resources, helping the global community accelerate progress in embodied intelligence, world modeling, robot learning, and real-world deployment.

AGIBOT WORLD 2026 Theme 2: Rich Interaction Dataset is available through the following open-source channels:

Hugging Face: huggingface.co/datasets/agibot-world/AgiBotWorld2026

Project Page: agibot-world.com/

For more about AGIBOT, visit AGIBOT.com or follow AGIBOT on:

https://x.com/AGIBOT_zhiyuan

https://www.linkedin.com/company/agibot/

https://www.youtube.com/@AGIBOT-robot

https://www.instagram.com/AGIBOT

https://www.tiktok.com/@agibot

https://www.facebook.com/AGIBOT.zhiyuan

About AGIBOT

AGIBOT is an Embodied AI foundation model company developing both the intelligence layer and the corresponding robotic embodiments needed to bring general intelligence into the physical world. AGIBOT’s “Three Intelligences in One” architecture integrates Locomotion Intelligence, Interaction Intelligence, and Manipulation Intelligence into a unified embodied system. Its portfolio spans humanoid robots, quadrupeds, dexterous systems, and commercial cleaning solutions. In March 2026, AGIBOT announced that its 10,000th robot had rolled off the production line.