aiqbench analysis
This paper describes an AI agent for the GLEE competition that pairs classical game-theoretic solvers (Rubinstein bargaining, Faratin concessions, Kamenica-Gentzkow disclosure) with an online Bayesian opponent model and a language model for messaging. Its most useful contribution is the reported gap between theoretically optimal policies and live play, along with the failure modes and guardrails the authors describe. The moral-framing result (χ² = 616.7 in 9,337 live messages) is a strong association in observational data, not evidence of causation, so readers should weigh it accordingly.
Game-Theoretic Solvers with Bayesian Opponent Modeling and Moral Rhetoric Framing for GLEE
arXiv:2610.37397v1 Announce Type: cross Abstract: We describe the agent we submitted to the GLEE Competition, which plays bargaining, negotiation, and persuasion games against a shared pool of human and AI opponents. The agent is built around three central components: a classical game-theoretic solver to compute numeric moves, an online Bayesian layer to estimate opponent archetypes, and a language model to write accompanying messages. We utilize Rubinstein bargaining for bargaining, Faratin concession for negotiation, and Kamenica--Gentzkow disclosure for persuasion, and adapt those models to live opponents using two separate layers: a Bayesian estimate of the opponent's hidden discount factor, reservation value, or signaling reliability, and a Thompson sampling bandit over five moral-foundation framings coupled with an evidence-intuition spectrum to establish message register. Most of the development effort went into the recurring problem that theoretically optimal policies perform poorly in certain live settings; we report associated failure modes, guardrails we built, and our hypotheses. We additionally evaluate our model over 13,200 Monte Carlo games against 11 synthetic opponent archetypes, an ablation of the opponent-modeling layer, and an extensive analysis of 9,337 live-game messages in which moral rhetoric framing in the persuasion setting is strongly correlated with whether the buyer purchased ($\chi^2=616.7$, $p \approx 3.7\times10^{-132}$), surviving controls for quality, round, and configuration.
Abstract:We describe the agent we submitted to the GLEE Competition, which plays bargaining, negotiation, and persuasion games against a shared pool of human and AI opponents. The agent is built around three central components: a classical game-theoretic solver to compute numeric moves, an online Bayesian layer to estimate opponent archetypes, and a language model to write accompanying messages. We utilize Rubinstein bargaining for bargaining, Faratin concession for negotiation, and Kamenica--Gentzkow disclosure for persuasion, and adapt those models to live opponents using two separate layers: a Bayesian estimate of the opponent's hidden discount factor, reservation value, or signaling reliability, and a Thompson sampling bandit over five moral-foundation framings coupled with an evidence-intuition spectrum to establish message register. Most of the development effort went into the recurring problem that theoretically optimal policies perform poorly in certain live settings; we report associated failure modes, guardrails we built, and our hypotheses. We additionally evaluate our model over 13,200 Monte Carlo games against 11 synthetic opponent archetypes, an ablation of the opponent-modeling layer, and an extensive analysis of 9,337 live-game messages in which moral rhetoric framing in the persuasion setting is strongly correlated with whether the buyer purchased ($\chi^2=616.7$, $p \approx 3.7\times10^{-132}$), surviving controls for quality, round, and configuration.
Computer Science and Game Theory (cs.GT); Computation and Language (cs.CL)
arXiv:2610.37397 [cs.GT]
(or arXiv:2610.37397v1 [cs.GT] for this version)
https://doi.org/10.48550/arXiv.2610.37397 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 305 words