AGIBOT WORLD CHALLENGE 2026 Advances Embodied AI Competition from Simulation to Real-Robot Testing at ICRA 2026

AGIBOT WORLD CHALLENGE 2026 marked a pivotal shift in embodied AI research by transitioning from simulation-based evaluations to real-robot testing. The event, held in Vienna as part of ICRA 2026, attracted 526 teams from 27 countries, including institutions like the Chinese Academy of Sciences, Tsinghua University, and companies such as Alibaba and vivo. Two tracks—Reasoning to Action (R2A) and World Model (WM)—focused on evaluating robots’ ability to understand tasks, plan actions, and model physical-world interactions. The Reasoning to Action track assessed robots’ full workflow, from environment perception to task execution, using the AGIBOT WORLD open-source dataset and Genie Sim 3.0. Teams like PrismBot from vivo and Shanghai RoboParty’s RP-VLA demonstrated advanced reasoning and manipulation skills, with PrismBot winning the championship. The competition also introduced a real-supermarket benchmark, where algorithms controlled robots in end-to-end retail tasks, including navigation, picking, and placement under physical constraints. The World Model track evaluated AI systems’ ability to predict physical-world changes and adapt to sensor inputs. NeoVerse-ABot, a collaboration between the Chinese Academy of Sciences and Amap CV Lab, emerged as the top performer. The challenge adopted a hybrid evaluation system, combining online automated testing with an offline real-robot final, ensuring standardized benchmarks and reproducible results. AGIBOT’s EWMBench and Genie Sim Benchmark provided a consistent framework for measuring embodied AI performance across simulations and real-world scenarios. By incorporating real-robot validation, the competition prioritized stability, adaptability, and long-horizon task reliability, aligning technical assessments with practical deployment needs. Over 100 teams surpassed the official baseline, reflecting strong global participation from academia and industry. The event underscored the growing emphasis on embodied AI’s real-world applicability, moving beyond theoretical benchmarks to tangible, deployable solutions. AGIBOT’s initiative highlights the industry’s push toward more robust and adaptable robotic systems capable of operating in dynamic environments.