English(EN) StalePO: Anchored Token-Level Preference Optimization using Legacy Post-Edits in Machine Translation

新的StalePO方法利用遗留数据改进机器翻译

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-16 04:00

研究人员开发了StalePO，这是一种新的优化方法，旨在改进在过时偏好数据上训练的机器翻译系统。传统方法在使用旧模型的后编辑时，可能会降低性能或无法纠正特定错误。StalePO通过确保两个响应的似然度均向下移动，将策略锚定到其基础响应，并在令牌级别应用KL约束来解决此问题。这种方法已显示出显著的收益，在英译印地语翻译中将质量检查提高了多达14.9个百分点，在英译土耳其语翻译中提高了4.6个百分点。 AI

影响通过有效利用遗留训练数据来提高机器翻译质量。

排序理由该集群包含一篇详细介绍机器翻译新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 14 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍机器翻译新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Rohit Dhaipule, Sukhdeep Singh Kharbanda, Prasanth Bathala, Pradyumna Lanka, Anubhav Shrimal · 2026-09-16 04:00

StalePO：使用机器翻译中的旧后编辑进行锚定令牌级偏好优化

arXiv:2609.16340v1 Announce Type: new Abstract: Machine translation systems are periodically upgraded to stronger models, but the available preference signal is human post-edits of an older system's outputs, which the newer model may already surpass. Moreover, collecting fresh po…

本文如何被排名

报道来源 [1]

StalePO：使用机器翻译中的旧后编辑进行锚定令牌级偏好优化

相关实体

相关话题