MOPD：面向LLM後訓練中能力融合的多教師同策略蒸餾

摘要

现代大语言模型在后训练阶段依赖强化学习来提升特定能力，但将多种能力整合至单一模型仍面临挑战。现有方法如离策略微调(Off-Policy Finetune)和混合强化学习(Mix-RL)存在效率低下或性能损失的问题。为此，我们提出多教师同策略蒸馏(MOPD)——一种用于融合多个领域RL教师能力的新型后训练范式：首先通过各领域专项RL训练获得一组领域教师，然后在学生模型的自生成轨迹上对这些教师进行蒸馏。该方法消除了暴露偏差并提供了密集的优化信号。在Qwen3-30B-A3B模型上，MOPD的性能超越了Mix-RL、级联RL、离策略微调和参数合并等基线方法，几乎完整继承了每位教师的全部能力。同时，MOPD支持各领域教师的并行独立开发，消除了多领域后训练中常见的跨领域耦合问题。该技术已应用于工业级前沿模型MiMo-V2-Flash的后训练阶段，充分验证了其在整合前沿大语言模型能力方面的实际价值。

English

Modern large language models (LLMs) rely on reinforcement learning during post-training to push specific capabilities, yet integrating multiple capabilities into one model remains hard. Existing methods, such as Off-Policy Finetune and Mix-RL, are either inefficient or lose performance. In this work, we propose Multi-teacher On-Policy Distillation (MOPD), a post-training paradigm for combining the capabilities of multiple domain RL teachers: we first run per-domain specialised RL to obtain a set of domain teachers, then distill these teachers into the student on its own rollouts. This eliminates exposure bias and provides a dense optimization signal. On Qwen3-30B-A3B, MOPD outperforms Mix-RL, Cascade RL, Off-Policy Finetune, and Param-Merge baselines, inheriting nearly all of each teacher's capability. MOPD also enables parallel, independent development of domain teachers, removing the cross-domain coupling typical of multi-domain post-training. MOPD has been deployed in the post-training of MiMo-V2-Flash, an industrial-scale frontier model, demonstrating its practical value for capability integration in frontier-scale LLMs.