Researchers have proposed a new approach to visual navigation called UniWM, which integrates planning and world modeling into a single framework. This allows agents to imagine future states and make decisions based on those predictions. The system uses a hierarchical memory mechanism to fuse short-term and long-term information, enabling stable and coherent reasoning over extended periods. Experiments show that UniWM improves navigation success rates by up to 30% compared to state-of-the-art systems. --- Why it matters: This matters because it addresses the limitations of current visual navigation systems, which often rely on modular designs that can lead to state-action misalignment and poor adaptability in novel or dynamic scenarios. The proposed approach has the potential to improve the robustness and generalizability of embodied agents in various applications. Source: https://arxiv.org/abs/2510.08713