Suite of Environments for RLSP Evaluation
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/HumanCompatibleAI/rlsp
下载链接
链接失效反馈官方服务：
资源简介：
该数据集提供了一系列环境，旨在通过为Alice和机器人提供真实的奖励、指定的奖励以及初始状态来评估RLSP算法。这些环境的设计目的是展示RLSP算法的特性，并评估其在推断奖励方面的工作性能。该任务涉及的是带有隐式偏好的强化学习。
This dataset provides a suite of environments designed to evaluate the RLSP algorithm by supplying Alice and the robot with ground-truth rewards, specified rewards, and initial states. These environments are constructed to demonstrate the characteristics of the RLSP algorithm and assess its performance in reward inference. The relevant task falls within the scope of reinforcement learning with implicit preferences.
提供机构：
Human Compatible AI
