Business World Model

Researchers have proposed a business world model (BWM) to enable intelligent systems to plan and execute business initiatives from high-level strategic objectives, and introduced WorldOlympiad, a benchmark for evaluating video-based world models.

The BWM concept, presented in a paper submitted in 2026^[2], aims to support autonomous decision-making by encoding business states, dynamics, constraints, objectives, and feasible action space. The proposed architecture integrates semantic data representations, probabilistic machine learning models, and deterministic business rules into a coherent structure for planning and counterfactual reasoning^[1]. Meanwhile, WorldOlympiad, also introduced in a 2026 paper^[2], is a benchmark that evaluates video-based world models across three dimensions: physical faithfulness, geometric consistency, and interaction fidelity. WorldOlympiad has 3 tracks: physical, geometry, and interaction^[2]. The physical track assesses whether generated videos follow interpretable rules using object segmentation and MLLM-as-judge. The geometry track evaluates structural consistency and camera-trajectory alignment by reconstructing generated videos with Gaussian splatting. The interaction track assesses whether generated rollouts follow complex action prompts and maintain smooth transitions.