WMT25通用机器翻译系统的初步排名
计算与语言
2025-08-26 v2
摘要
我们公布了提交至WMT25通用机器翻译共享任务的机器翻译（MT）系统的初步排名，该排名由自动评估指标确定。由于这些排名源自自动评估，它们可能对采用重排序技术（如质量估计或最小贝叶斯风险解码）的系统表现出偏好。官方WMT25排名将基于人工评估，其更为可靠并将取代这些结果。官方WMT25排名将基于人工评估，其更为可靠并将取代这些结果。现在发布这些结果的目的是协助任务参与者撰写其系统描述论文，而非提供最终结论。
引用
@article{arxiv.2508.14909,
title = {Preliminary Ranking of WMT25 General Machine Translation Systems},
author = {Tom Kocmi and Eleftherios Avramidis and Rachel Bawden and Ondřej Bojar and Konstantin Dranch and Anton Dvorkovich and Sergey Dukanov and Natalia Fedorova and Mark Fishel and Markus Freitag and Thamme Gowda and Roman Grundkiewicz and Barry Haddow and Marzena Karpinska and Philipp Koehn and Howard Lakougna and Jessica Lundin and Kenton Murray and Masaaki Nagata and Stefano Perrella and Lorenzo Proietti and Martin Popel and Maja Popović and Parker Riley and Mariya Shmatova and Steinþór Steingrímsson and Lisa Yankovskaya and Vilém Zouhar},
journal= {arXiv preprint arXiv:2508.14909},
year = {2025}
}