MSQA: A Natively Sourced Multilingual and Multicultural SimpleQA Benchmark
MSQA:一个原生来源的多语言多文化SimpleQA基准
机构 * M-A-P ; ByteDance Seed(字节跳动Seed) ; Beijing University of Posts and Telecommunications(北京邮电大学) ; Nanjing University(南京大学)
AI总结 提出MSQA基准,包含1064个原生问题覆盖11种语言和5个文化维度,评估18个LLM发现文化能力随预训练暴露程度下降,且推理时补救措施无效。
Comments Due to the company's data approval issue, we need to withdraw the article