Year: 2026
Type: preprint
Abstract: Test-time reasoning has significantly improved performance in domains ranging from games to language models. However, test-time policy changes with formal guarantees on the performance of the resulting strategy remain a challenge in two-player zero-sum imperfect-information games. Existing solutions are limited to tabular methods or single gradient step updates. In this work, we investigate policy-gradient algorithms as a method for scalable test-time reasoning. We extend the concept of gadget game, tabular technique for test-time search, to the reinforcement learning setting. Unlike prior approaches, we represent the gadget game implicitly by modified sampling and neural policy rather then explicitly by constructing it, thereby removing constraints on subgame size. Furthermore, we formally prove that, unlike prior tabular algorithms, regularized policy-gradient algorithms limit possible strategy degradation caused by test-time reasoning, even without the gadget games. Our evaluation across small- and large-scale games confirms that additional test-time training often substantially improves performance relative to the blueprint strategy.
Source: arXiv (Cornell University)
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: G12500569, RI: Medium: Techniques for Massive-Scale Strategic Reasoning: Imperfect-Information Subgame Solving and Offering Guarantees in Simulation-Based Games, G1471787206, Cryptic Speciation in Acanthocyclops Vernalis Complex (Crustacea: Copepoda): the Hypothesized Role of Chromatin Diminution, G2291329386, G2759202153, G3524244961, G4131035492, Advancing Large Language Model Agents through Game Playing, G6282746026, G8470341104, G8876996369, G937205589
OpenAlex ID: W7204900132
Cites:
Cited by:
Related to: