# 开源大语言模型多语言机器翻译的免参考后训练

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-08-11 08:00
- AIHOT 分数：55
- AIHOT 链接：https://aihot.virxact.com/items/cmspmcnwa04bcrojexo17s4k8
- 原文链接：https://arxiv.org/abs/2608.10812

## AI 摘要

研究团队基于监督微调的MiLMMT-46-v0.1，采用GRPO与免参考质量评估奖励，并通过SFT与RL检查点线性插值得到MiLMMT-46-v1.0。该模型在46种语言上持续优于SFT版本，超越Seed-X、HY-MT2、TranslateGemma等开源基线，免参考评分领先Google Translate、Gemini 3 Pro、GPT-5等专有系统。模型与代码已开源。

## 正文

We study reference-free post-training for multilingual machine translation with open large language models. Starting from the supervised-finetuned MiLMMT-46-v0.1 models, we apply Group Relative Policy Optimization (GRPO) with a reward that averages two reference-free quality estimation models and is gated by language identification. We then linearly interpolate the supervised fine-tuning (SFT) and reinforcement learning (RL) model checkpoints to obtain MiLMMT-46-v1.0. Across 46 languages, the resulting models consistently improve translation quality over their SFT counterparts, outperform strong recent open baselines, including Seed-X, HY-MT2, and TranslateGemma, and achieve leading reference-free scores against evaluated proprietary systems such as Google Translate, Gemini 3 Pro, and GPT-5. We further investigate on-policy distillation and find that it reaches, but does not surpass, the quality frontier achieved by RL with checkpoint interpolation. We release the models and code to facilitate future research.