GIANT: General intelligent AgeNt trainerŠmid, Marko (Avtor) Ravber, Miha (Avtor) machine learningevolutionary algorithmsgenetic programmingoptimizationgamesmulti-agent systemsIntelligent agent training is an active research area in artificial intelligence, primarily driven by reinforcement learning methods that utilize neural networks as decision-making systems. To broaden this field of research, we present the General Intelligent AgeNt Trainer, a flexible, modular, and scalable platform for training intelligent agents. The current implementation integrates evolutionary algorithms, with a particular focus on genetic pro gramming and behavior trees as interpretable decision-making systems. The platform’s architecture is designed for extensibility, allowing integration of additional machine learning techniques, optimization strategies, and decision-making systems. Furthermore, the platform aims to establish a benchmark for evaluating and comparing machine learning algorithms in both single-agent and multi-agent environments.Elsevier B.V.20262026-03-23 14:17:41Članek v reviji97592sl