Reinforcement Learning (RL)-Based Recommender Systems (RSs) are increasingly recognized for their ability to improve long-term user engagement. Yet, the field grapples with challenges such as the absence of accessible frameworks, inconsistent evaluation standards, and the complexity of replicating prior work. Addressing these obstacles, we present EasyRL4Rec, a user-friendly and efficient library tailored for RL-based RSs. EasyRL4Rec features lightweight, diverse RL environments built on five widely-used public datasets, and is equipped with comprehensive core modules that offer rich options to ease the development of models. It establishes consistent evaluation criteria with a focus on long-term impacts and introduces customized solutions for state modeling and action representation tailored to recommender systems. Additionally, we share valuable insights gained from extensive experiments with current methods. EasyRL4Rec aims to facilitate the model development and experimental process in the domain of RL-based RSs. The library is openly accessible at https://github.com/chongminggao/EasyRL4Rec.

翻译：基于强化学习的推荐系统因其提升长期用户参与度的能力而日益受到认可。然而，该领域仍面临缺乏易用框架、评估标准不一致以及复现先前工作复杂性等挑战。针对这些问题，我们提出EasyRL4Rec——一个专为基于强化学习的推荐系统量身定制的高效易用库。EasyRL4Rec基于五个广泛使用的公开数据集构建轻量级、多样化的强化学习环境，并配备提供丰富选项的综合核心模块以简化模型开发。该库建立以长期影响为重点的标准化评估准则，并引入针对推荐系统的状态建模与动作表示定制化解决方案。此外，我们分享通过大量实验获得的关于现有方法的宝贵洞察。EasyRL4Rec旨在促进基于强化学习的推荐系统领域的模型开发与实验流程。该库已在https://github.com/chongminggao/EasyRL4Rec 开放访问。