aiqbench analysis
This arXiv study tests whether LLMs' cooperation in an iterated prisoner's dilemma holds when only the payoff scale or prompt language changes. It finds that it doesn't: cooperation shifts with the numerical scale of payoffs, the effect varies across models and languages, and persona effects reverse at larger stakes. For readers deploying LLM agents, the takeaway is that behavioural benchmarks are sensitive to evaluation setup, so cooperation results should not be treated as stable model traits without testing across scales, languages, and models.
Payoff scaling shapes cooperation in LLM agents across languages
arXiv:2601.19082v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly deployed as autonomous agents that negotiate, coordinate, and make decisions on behalf of human users. Understanding how they cooperate is therefore important for AI governance and machine behaviour research. We examine two factors: payoff magnitude and interaction language. Classical game theory predicts that multiplying all payoffs by a positive constant should not change behaviour because the underlying strategic incentives remain unchanged. We test this prediction using five widely adopted commercial LLMs playing an iterated prisoner's dilemma across ten payoff scales and five languages (English, French, Vietnamese, Chinese and Arabic). Contrary to theoretical expectations, changing only the numerical scale of payoffs substantially alters cooperation. Models respond differently, often in opposite directions, meaning that averages across models can conceal important behavioural heterogeneity. Sensitivity to payoff scale also depends strongly on prompt language. Moreover, payoff scale modifies the effects of persona instructions: at sub-unit stakes, where no payoff shown exceeds 1, all five models cooperate more under a selfish than a cooperative persona, whereas Claude and Gemini reverse this pattern at larger scales. These findings show that cooperation measures depend critically on payoff scale and language and should not be treated as fixed model traits.
Abstract:Large language models (LLMs) are increasingly deployed as autonomous agents that negotiate, coordinate, and make decisions on behalf of human users. Understanding how they cooperate is therefore important for AI governance and machine behaviour research. We examine two factors: payoff magnitude and interaction language. Classical game theory predicts that multiplying all payoffs by a positive constant should not change behaviour because the underlying strategic incentives remain unchanged. We test this prediction using five widely adopted commercial LLMs playing an iterated prisoner's dilemma across ten payoff scales and five languages (English, French, Vietnamese, Chinese and Arabic). Contrary to theoretical expectations, changing only the numerical scale of payoffs substantially alters cooperation. Models respond differently, often in opposite directions, meaning that averages across models can conceal important behavioural heterogeneity. Sensitivity to payoff scale also depends strongly on prompt language. Moreover, payoff scale modifies the effects of persona instructions: at sub-unit stakes, where no payoff shown exceeds 1, all five models cooperate more under a selfish than a cooperative persona, whereas Claude and Gemini reverse this pattern at larger scales. These findings show that cooperation measures depend critically on payoff scale and language and should not be treated as fixed model traits.
Accepted in Interface Focus (Royal Society). 57 pages, 17 figures, 25 tables, including supplementary material
Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
arXiv:2601.19082 [cs.AI]
(or arXiv:2601.19082v3 [cs.AI] for this version)
Full text under the publisher's open licence · 240 words
Related coverage
Sycophancy Undermines Epistemic Vigilance in Cooperative Vision-Language Tasks
arXiv · Aug 3, 2026
Multi-Agentic AI for Fairness-Aware and Accelerated Multi-modal Large Model Inference in Real-world Mobile Edge Networks
arXiv · Oct 27, 2026
Training Language Models to Cooperate with Inference-Time Controllers
arXiv · Jul 28, 2026
CoopReflect: Towards Natural Language Communication for Cooperative Autonomous Driving via Multi-Agent Learning
arXiv · Jul 28, 2026