UniWAM: Unified World-Action Model

UniWAM is a novel architecture that integrates a physical reasoner, world generator, and action predictor to jointly learn semantic understanding, visual generation, and action prediction. The model leverages human and robot data, with a focus on data cleaning, natural language action representation, and post-training techniques like future visual noise augmentation and history-conditioned flow matching. It achieves state-of-the-art performance across multiple benchmarks, including robustness, generalization, and long-horizon task execution.