Context-dependent agent evaluation with orthogonal equilibrium learning

2026-09-25

AI Summary

This study addresses the challenge that agent evaluation under offline feedback struggles to reflect heterogeneous collective preferences, and proposes the NashEval framework. By integrating social choice theory with game theory, this method formulates evaluation as a contextual game. It jointly learns Nash equilibria through debiased estimation and orthogonal loss functions, thereby circumventing context-by-context resolution. Furthermore, it theoretically proves that estimation errors affect equilibrium risk only at higher orders. Experimental results demonstrate that NashEval significantly enhances the robustness of equilibrium learning, consistently and accurately identifying the set of winning agents across diverse contexts.

Abstract

Many applications require to evaluate agents under contextual information (e.g., a prompt, task, or user group). We study how to perform such context-dependent agent evaluation from offline feedback. Existing score-based models for this purpose (e.g., Bradley-Terry) impose a transitive preference ordering, which fails to reflect collective preferences when human judgements are heterogeneous. Inspired by social choice theory, we frame evaluation as a contextual game between two players, each selecting a distribution over agents as the strategy to receive greater collective preference than the other. Then, the support of the Nash equilibrium defines a context-specific set of winners. However, learning context-specific equilibria from offline logs is difficult because each context reveals human feedback on only a subset of agents, and, hence, a naive plug-in estimator can therefore be biased. To address these challenges, we propose NashEval, a general framework for robust contextual equilibrium learning. NashEval first constructs debiased estimates of the contextual payoff matrix that characterizes the game. NashEval then learns the context-to-equilibrium mapping with a tailored orthogonal loss, which avoids the need to solve a separate game for each context. We show theoretically that errors in estimating the nuisance functions underlying the payoff matrix affect the risk of the learned equilibrium (i.e., exploitability) only through higher-order terms. Across various experiments, NashEval improves robustness of equilibrium learning and consistently identifies the set of top-performing agents across contexts.

Problem

• context-dependent agent evaluation
• offline feedback
• Nash equilibrium
• heterogeneous preferences
• debiased estimation

Innovation

• Context-dependent agent evaluation
• Nash equilibrium learning
• Orthogonal loss
• Debiased estimation
• Social choice theory