Search papers, labs, and topics across Lattice.
The paper introduces the World-Value-Action (WAV) model, a novel framework for Vision-Language-Action (VLA) systems that enables implicit planning by learning a structured latent representation of future trajectories conditioned on visual observations and language instructions. This approach uses a world model to predict future states and a trajectory value function to evaluate their long-horizon utility, formulating action generation as inference in the latent space. Experiments show that WAV outperforms state-of-the-art methods in long-horizon and compositional tasks due to its ability to concentrate probability mass on high-value and dynamically feasible trajectories.
Planning in latent space, rather than directly in action space, overcomes the exponential decay in feasible trajectories, unlocking efficient long-horizon decision making for embodied agents.
Vision-Language-Action (VLA) models have emerged as a promising paradigm for building embodied agents that ground perception and language into action. However, most existing approaches rely on direct action prediction, lacking the ability to reason over long-horizon trajectories and evaluate their consequences, which limits performance in complex decision-making tasks. In this work, we introduce World-Value-Action (WAV) model, a unified framework that enables implicit planning in VLA systems. Rather than performing explicit trajectory optimization, WAV model learn a structured latent representation of future trajectories conditioned on visual observations and language instructions. A learned world model predicts future states, while a trajectory value function evaluates their long-horizon utility. Action generation is then formulated as inference in this latent space, where the model progressively concentrates probability mass on high-value and dynamically feasible trajectories. We provide a theoretical perspective showing that planning directly in action space suffers from an exponential decay in the probability of feasible trajectories as the horizon increases. In contrast, latent-space inference reshapes the search distribution toward feasible regions, enabling efficient long-horizon decision making. Extensive simulations and real-world experiments demonstrate that the WAV model consistently outperforms state-of-the-art methods, achieving significant improvements in task success rate, generalization ability, and robustness, especially in long-horizon and compositional scenarios.
