Cognitively Constrained Partners for Zero-Shot Coordination in Real-Time Partitioned Control

Computer Science > Machine Learning arXiv:2610.31726 (cs) [Submitted on 22 Oct 2026] Title: Cognitively Constrained Partners for Zero-Shot Coordination in Real-Time Partitioned Control Authors: Ash Simmons, George De Ath, Richard M. Everson.

计算机科学 > 机器学习 arXiv:2610.31726 (cs) [提交于 2026 年 10 月 22 日] 标题：用于实时分区控制中零样本协作的认知约束伙伴 作者：Ash Simmons, George De Ath, Richard M. Everson。

Abstract: Effective human-machine teaming requires agents that coordinate with unpredictable human partners. Reinforcement learning from live human feedback or large behavioural datasets is expensive, while purely simulated training overfits to artificial conventions and fails at Zero-Shot Coordination with real people.

摘要：有效的人机协作需要能够与不可预测的人类伙伴进行协调的智能体。通过实时人类反馈或大型行为数据集进行强化学习成本高昂，而纯模拟训练则会过拟合于人为设定的惯例，且在与真实人类进行零样本协作时表现不佳。

We propose a framework for training teaming agents against Cognitively Constrained Partners: simulated partners whose diversity emerges from procedural constraints grounded in human physiological literature, specifically reaction delays and perceptual limitations, rather than from reward shaping or historical training checkpoints.

我们提出了一种针对“认知约束伙伴”训练协作智能体的框架：这些模拟伙伴的多样性源于基于人类生理学文献的程序性约束（特别是反应延迟和感知局限性），而非源于奖励塑造或历史训练检查点。

In a modified Lunar Lander environment, we report preliminary evidence that these partners play more like real people than unconstrained agents do, and that agents trained against them match the strongest cross-play baselines while being the only ones whose performance does not depend on how their training partners were generated.

在一个改进的月球着陆器环境中，我们报告了初步证据，表明这些伙伴比无约束的智能体表现得更像真实人类，并且针对它们训练出的智能体能够匹配最强的交叉博弈基准，同时也是唯一在性能上不依赖于其训练伙伴生成方式的智能体。

We also describe a human-participant study designed so that its rankings can be compared directly against cross-play, testing how well that standard evaluation predicts real human-agent teaming.

我们还描述了一项人类参与者研究，其设计旨在使研究排名能够直接与交叉博弈进行比较，从而测试该标准评估在预测真实人机协作方面的效果。