用强化学习生成难译文本，自动测试机器翻译短板。
Generating Adversarial Texts for Machine Translation via GRPO
- 基于GRPO强化学习重写文本，提升翻译难度。
- WMT25上翻译质量从0.63降至0.48，保持语义和语法。
- 适用于评估模型鲁棒性，适合研究者与评测人员。
随着机器翻译系统持续进步，传统基准测试难以揭示其潜在弱点。传统挑战集构建依赖昂贵的人工设计，而自动化方法难以同时保证翻译难度与语言多样性。本文提出一种可扩展的强化学习方法，将现有源文本重写为对机器翻译更具挑战性的形式。通过组相对策略优化（GRPO）微调大语言模型，利用翻译难度作为奖励信号，并施加语义相似性、语法正确性和长度近似保持的约束。在WMT25数据集上，该方法将平均COMET得分从0.63显著降至0.48，而基础模型保持在0.64；在未见的WMT19-WMT24基准上验证了良好泛化能力。人工评估显示，重写文本大幅降低翻译质量，自然度略有下降，语法几乎不变。代码已公开以支持可复现性。
原文摘要 · Abstract (English)
As machine translation (MT) systems continue to improve, standard benchmarks become less informative for exposing remaining weaknesses. Traditional methods for creating challenging test sets rely on expensive manual creation or curation, while automated approaches struggle to produce sets with the necessary translation difficulty and linguistic diversity. We propose a scalable reinforcement-learning-based approach for rewriting existing source texts into instances that are more difficult to translate for MT systems. We fine-tune a large language model with Group Relative Policy Optimization (GRPO), using reward signals based on translation difficulty together with constraints for semantic similarity, grammaticality, and approximate length preservation. On WMT25, our approach substantially reduces average COMET translation quality from 0.63 to 0.48, while preserving grammaticality and readability, whereas the base model remains at 0.64. Evaluations on the unseen WMT19-WMT24 benchmarks confirm that this behavior generalizes beyond the training data, and human evaluation further shows that the rewrites substantially lower translation quality while incurring a moderate drop in naturalness and only a small change in grammaticality. We release our code to support reproducibility.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。