[2603.27751] SkyNet: Belief-Aware Planning for Partially-Observable Stochastic Games

Computer Science > Artificial Intelligence arXiv:2603.27751 (cs) [Submitted on 29 Mar 2026] Title:SkyNet: Belief-Aware Planning for Partially-Observable Stochastic Games Authors:Adam Haile View a PDF of the paper titled SkyNet: Belief-Aware Planning for Partially-Observable Stochastic Games, by Adam Haile View PDF HTML (experimental) Abstract:In 2019, Google DeepMind released MuZero, a model-based reinforcement learning method that achieves strong results in perfect-information games by combining learned dynamics models with Monte Carlo Tree Search (MCTS). However, comparatively little work has extended MuZero to partially observable, stochastic, multi-player environments, where agents must act under uncertainty about hidden state. Such settings arise not only in card games but in domains such as autonomous negotiation, financial trading, and multi-agent robotics. In the absence of explicit belief modeling, MuZero's latent encoding has no dedicated mechanism for representing uncertainty over unobserved variables. To address this, we introduce SkyNet (Belief-Aware MuZero), which adds ego-conditioned auxiliary heads for winner prediction and rank estimation to the standard MuZero architecture. These objectives encourage the latent state to retain information predictive of outcomes under partial observability, without requiring explicit belief-state tracking or changes to the search algorithm. We evaluate SkyNet on Skyjo, a partially observable, non-zero-sum, stochastic card ...