大模型训练时表现好，上线却变差？
大模型用强化学习微调时，常出现训练时效果不错、上线后却变差的情况。研究者发现，问题出在训练和推理使用了不同的引擎——即使模型参数同步，同一段文本在两个引擎中算出的概率也不一样，导致训练时优化的策略在推理时并不成立。他们提出一个新框架，在训练中额外引入一个“推理侧差距代理”，只接受那些在推理侧也能保证提升的更新。实验表明，这能显著提升推理性能和训练稳定性。它不是你明天能用上的，但解释了为什么大模型微调容易“翻车”，并给出了一个更可靠的优化方向。
📄 原文摘要(英文)
Reinforcement learning (RL) has gained growing attention in large language model (LLM) post-training, yet RL training remains fragile and can suffer from instability or collapse. One vital cause is training-inference mismatch: LLM adopts separate inference and training engines for generation efficiency and training precision, which in practice exhibits inconsistent probabilities for the same trajectories on training and inference sides, even with synchronized model parameters. This naturally induces a special type of off-policyness ever existing and poisoning the training. Prior works have made various efforts in addressing the off-policyness to stabilize the training policies under the mismatch. In this paper, we point out the objective misalignment neglected by existing works that an effective update to the policy in the training engine not necessarily ensures the improvement of the inference policy, i.e., the one used in deployment. To this end, we propose a new policy optimization objective for LLM RL, named Monotonic Inference Policy Improvement (MIPI). Following this principle, we introduce Monotonic Inference Policy Update (MIPU), a two-step LLM RL framework that constructs sampler-referenced candidate updates and selectively accepts synchronized candidates using an inference-side gap proxy. Experiments conducted on two model scales under high mismatch show that MIPU improves average reasoning performance and training stability.