McGill-NLP/agent-reward-bench
收藏Hugging Face2025-04-21 更新2025-05-31 收录
下载链接:
https://hf-mirror.com/datasets/McGill-NLP/agent-reward-bench
下载链接
链接失效反馈官方服务:
资源简介:
AgentRewardBench是一个用于评估网络代理轨迹自动评价的基准数据集。它包含了多个代理在不同网络环境中的行为轨迹,以及人类评判者对这些轨迹的评价。数据集适用于机器人学任务,并提供了不同大小的数据子集。
AgentRewardBench is a benchmark dataset for evaluating automatic evaluations of web agent trajectories. It includes behaviors of multiple agents in various web environments, as well as evaluations of these trajectories by human judges. The dataset is suitable for robotics tasks and provides subsets of different sizes.
提供机构:
McGill-NLP



