XRL-Bench

XRL-Bench是由伏羲人工智能实验室开发的一个标准化基准，用于评估和比较可解释强化学习（XRL）技术。该数据集支持表格和图像数据输入，用于状态解释，并包含三个主要模块：标准RL环境、基于状态重要性的解释器和标准评估器。XRL-Bench旨在为XRL方法的持续发展和评估提供坚实基础，特别适用于在线游戏服务等实际应用场景，帮助理解和优化强化学习模型的决策过程。

XRL-Bench is a standardized benchmark developed by Fuxi AI Lab for evaluating and comparing Explainable Reinforcement Learning (XRL) techniques. This benchmark supports tabular and image data inputs for state explanation, and includes three core modules: standard RL environments, state importance-based interpreters, and standard evaluators. XRL-Bench aims to provide a solid foundation for the continuous development and evaluation of XRL methods, and is particularly applicable to real-world scenarios such as online game services, helping researchers understand and optimize the decision-making processes of reinforcement learning models.

搜集汇总

背景与挑战

背景概述

XRL-Bench是由伏羲人工智能实验室开发的可解释强化学习（XRL）评估基准，支持表格和图像数据输入，包含标准RL环境、解释器和评估器三个模块，旨在促进XRL方法的发展并应用于在线游戏等实际场景，以优化决策过程理解。