Context-based evaluation finds top agents with robust game approach
Context-dependent agent evaluation with orthogonal equilibrium learning
Artificial Intelligence
Summary
Evaluating how good different agents are can depend on the situation, like a task or user group. The authors look at ways to judge agents based on past feedback, but old methods assume that preferences are straightforward and consistent, which is not always true when people disagree. They treat evaluation as a game between two players choosing agents to see which group is preferred overall based on context. Their method, called NashEval, estimates this game carefully from limited data and learns which agents perform best in each situation, even with noisy or incomplete feedback.
What this means in practice
- •For machine learning engineers: Improve model ranking under different user contexts using debiased equilibrium evaluation from offline data.
- •For recommendation system teams: Identify top recommendations among diverse options for varied user groups by learning context-specific winners robust to feedback biases.
Authors
Haorui Ma, Zehua Zang, Jiangmeng Li, Yi Li, Fanjing Xu, Stefan Feuerriegel
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.