AGIBOT announced on 22 Oct 2026 the open-source release of its “Physical Reasoning” dataset, the fourth release of AGIBOT WORLD 2026. The files sit on the AGIBOT WORLD 2026 dataset page on Hugging Face, which has a PhysicalReasoning folder and was last updated on 24 Oct 2026.
AGIBOT says the release covers “over 10 categories of logical reasoning scenarios” and nine reasoning abilities, among them long-horizon planning and subtask ordering. Tasks are sorted into three difficulty levels: Easy, Medium and Hard. The dataset provides high-level question-and-answer data together with low-level action trajectories and annotates execution outcomes; successful and erroneous samples are given equal weight, according to the release.
The release does not give a sample count or a data size. AGIBOT says the dataset “has already been used to support AGIBOT’s upcoming agent framework and to train its next-generation foundation model.”