Research map / Learning theory and algorithms / Submodular and game algorithms

Stabilizing RL+Search for Imperfect-Information Extensive-Form Games

NeurIPS 2026 · Learning theory and algorithms › Submodular and game algorithms · approach: Zero-sum game learning dynamics

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Identifies a second source of training fragility in ReBeL-style RL+Search for imperfect-information extensive-form games: the search budget is fixed across all subgames, creating a rigid coupling between subgame search and training.

Problem

RL+Search training for imperfect-information extensive-form games (as in ReBeL) is fragile; ReBeL already samples the pivot iteration uniformly from the iteration budget, but the search budget is still held fixed across all subgames.

Approach

Proposes addressing the fixed per-subgame search budget, which rigidly couples subgames, to stabilize RL+Search training, building on ReBeL's uniform pivot-iteration sampling. Abstract is truncated, so the exact mechanism is not stated.

Experimental setup

Builds on: ReBeL

Closest papers by meaning

• Reevaluating Policy Gradient Methods for Imperfect-Information Games (ICLR 2026 · similarity 0.88)
• RL-CFR: Improving Action Abstraction for Imperfect Information Extensive-Form Games with Reinforcement Learning (ICML 2024 · similarity 0.88)
• Combining Deep Reinforcement Learning and Search for Imperfect-Information Games (NeurIPS 2020 · similarity 0.88)
• From Average-Iterate to Last-Iterate Convergence in Games: A Reduction and Its Applications (NeurIPS 2025 · similarity 0.87)
• Local and Adaptive Mirror Descents in Extensive-Form Games (NeurIPS 2024 · similarity 0.87)
• The Effective Horizon Explains Deep RL Performance in Stochastic Environments (ICLR 2024 · similarity 0.87)