AI学会从“未来的自己”身上偷师
大模型训练通常需要人类反馈或大量数据，但这篇论文让AI自己教自己——而且是从“未来的自己”那里学。研究者针对扩散语言模型（一种非逐字生成、而是整体“显影”出文本的模型）设计了一套新方法：让模型先生成完整答案，再回头用这个答案作为“未来经验”去指导自己优化推理过程。在四个推理测试中，它只用强化学习十分之一的训练步数就超越了传统方法。这不是你明天能直接用的工具，但它指向一个趋势：AI可能不再需要那么多人类标注，而是靠自我迭代变得更聪明。
📄 原文摘要(英文)
On-policy self-distillation (OPSD) has proven effective for post-training large language models (LLMs), yet its application to diffusion LLMs (dLLMs) remains unexplored. Existing OPSD methods are inherently autoregressive-centric. They inject privileged information via left-to-right prefix conditioning with token-level divergence supervision, a design that fundamentally conflicts with the arbitraryorder generation of dLLMs. We introduce d-OPSD, the first OPSD framework tailored for dLLMs. Our approach makes two core contributions. First, we reframe self-teacher construction by using self-generated answers as suffix conditioning, enabling the student model to learn from "self future-experience" rather than privileged prefixes. Second, we shift supervision from token-level to step-level, aligning training with the iterative denoising process of dLLMs. Experiments across four reasoning benchmarks show that d-OPSD consistently outperforms RLVR and SFT baselines with superior sample efficiency, requiring only around 10% of the optimization steps by RLVR and opening a promising pathway for dLLM posttraining. The code is available at https://github.com/xingzhejun/d-OPSD.