ibm-research/justrank_judge_scores
收藏Hugging Face2025-06-08 更新2025-05-31 收录
下载链接：
https://hf-mirror.com/datasets/ibm-research/justrank_judge_scores
下载链接
链接失效反馈官方服务：
资源简介：
JuStRank Judge Scores数据集是关于LLM评估者和奖励模型对63个系统在Arena Hard v0.1基准的500个问题所给出的质量评分的原始数据集。这些评分用于评估LLM评估者在系统级排名中的性能和行为，并用于创建JuStRank排行榜。
The JuStRank Judge Scores dataset consists of quality scores given by LLM judges and reward models for the outputs of 63 systems on 500 questions from the Arena Hard v0.1 benchmark. These scores are used to evaluate the performance and behavior of LLM judges as system-level rankers, and are also used to create the JuStRank Leaderboard.
提供机构：
ibm-research
