Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning

Christo Mathew; Wentian Wang; Jacob Feldman; Lazaros K. Gallos; Paul B. Kantor; Vladimir Menkov; Hao Wang

Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning

机器学习 2025-10-24 v3 人工智能机器学习

作者: Christo Mathew , Wentian Wang , Jacob Feldman , Lazaros K. Gallos , Paul B. Kantor , Vladimir Menkov , Hao Wang

摘要

我们探讨了在 Game Of Hidden Rules (GOHR) 环境中的强化学习，该环境是一个复杂的谜题，要求智能体推断并执行隐藏规则，以将 6 $\times$ 6 棋盘上的棋子放入桶中。我们研究了两种状态表示策略，即基于特征的 (Feature-Centric, FC) 和基于对象的 (Object-Centric, OC)，并采用基于 Transformer 的 Advantage Actor-Critic (A2C) 算法进行训练。该代理仅获得部分观察，必须通过经验同时推断主导规则并学习最优策略。我们在多种基于规则和基于试验列表的实验设置中评估了模型，分析了迁移效应和表示对学习效率的影响。

关键词

强化学习隐藏规则 Transformer A2C 状态表示

引用

@article{arxiv.2509.06213,
  title  = {Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning},
  author = {Christo Mathew and Wentian Wang and Jacob Feldman and Lazaros K. Gallos and Paul B. Kantor and Vladimir Menkov and Hao Wang},
  journal= {arXiv preprint arXiv:2509.06213},
  year   = {2025}
}

Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning

摘要

关键词

引用

相关论文