CohereLabsCommunity/multilingual-reward-bench
收藏Hugging Face2025-07-23 更新2025-05-31 收录
下载链接：
https://hf-mirror.com/datasets/CohereLabsCommunity/multilingual-reward-bench
下载链接
链接失效反馈官方服务：
资源简介：
多语言奖励基准 (M-RewardBench) 数据集是一个用于评估奖励模型在多语言环境下表现的基准数据集。它包含来自 RewardBench 的约 2.87k 文本样本，翻译成 23 种其他语言。数据集包括提示-选择-拒绝偏好三元组，涵盖了通用功能和多语言知识。该数据集由 Aya RM 多语言团队整理，并获得了 Cohere 的研究计算资助。
The Multilingual Reward Bench (M-RewardBench) dataset is a benchmark for evaluating reward models in multilingual settings. It includes about 2.87k text samples from RewardBench translated into 23 other languages. The dataset consists of prompt-chosen-rejected preference triples and covers general-purpose capabilities and multilingual knowledge. It was curated by the Aya RM Multilingual Team and funded by Coheres Research Compute Grant.
提供机构：
CohereLabsCommunity
