HiTZ/es_eu-recon

该数据集是用于论文《Towards Reliable Multilingual Judge Models: An Empirical Study》的基准数据集，旨在支持多语言评估模型的可靠性研究。它源自原始英语数据集RECON，并提供了多个变体：recon_es（所有字段翻译为西班牙语）、recon_eu（所有字段翻译为巴斯克语）、recon_io_es（仅模型输入和输出翻译为西班牙语，其他字段保留英语）和recon_io_eu（仅模型输入和输出翻译为巴斯克语，其他字段保留英语）。这些变体用于评估模型在不同语言环境下的性能。This dataset is a benchmark used in the paper Towards Reliable Multilingual Judge Models: An Empirical Study, designed to support research on the reliability of multilingual evaluation models. It is derived from the original English dataset RECON and includes several variants: recon_es (all fields translated into Spanish), recon_eu (all fields translated into Basque), recon_io_es (only the model input and output translated into Spanish, with other fields remaining in English), and recon_io_eu (only the model input and output translated into Basque, with other fields remaining in English). These variants are used to assess model performance across different linguistic contexts.