Reference-Free Post-Training of Open Large Language Models for Multilingual Machine Translation

Here's a thinking process: 1. **Analyze User Request:** - Role: Technical news summarizer - Task: Condense provided news into brief HTML summary - Input: Title, source, URL, author, date, description - Output format: HTML with `

` containing 2-4 sentence summary, and `Read original` - Rules: Concise (2-4 sentences), precise technical language, no invented info, output ONLY valid HTML, English only - News to summarize provided 2. **Analyze the News Content:** - Title: "Reference-Free Post-Training of Open Large Language Models for Multilingual Machine Translation" - Source: huggingface/daily-papers - URL: https://arxiv.org/abs/2608.10812 - Author: Chris Han, Pengzhi Gao, Pei Fu, Jian Luan - Date: 2026-08-10 - Description: "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