[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-opdlm-on-policy-distillation-dllm-1-7000x":3,"news-related-b45f5c46-c982-4410-9452-07a9f779218f":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"b45f5c46-c982-4410-9452-07a9f779218f","On-Policy Distillation 把扩散语言模型训练成本砍到 1\u002F15~1\u002F7000","扩散语言模型（DLM）以并行解码获得速度优势，但训练成本一直是工程化的硬伤。arXiv:2606.06712（v1, 2026-06-08）从「ARLM 如何平滑过渡到 DLM」的工程视角切入，提出一种**自蒸馏 + 在线策略**的转换范式，把 DLM 预训练从「重训一遍」拉回「ARLM 后训练」的范畴。\n\n## 核心思路\n\n传统做法是直接拿 ARLM 改双向注意力，再套 DLM 目标重新预训练。作者指出这会触发两类分布漂移：**目标漂移**（从 next-token prediction 切到随机掩码预测会丢失世界知识）和**轨迹漂移**（训练走随机掩码，推理走基于置信度的解码，二者不一致）。\n\nOPDLM 用 **self-On-Policy Distillation** 解决：学生（带双向注意力的 ARLM）自己生成轨迹，原始冻结的 ARLM 作为教师在同一条轨迹上提供目标 logits。学生被直接拉到「推理时实际会走」的分布上。\n\n## 效果\n\n论文报告在多种任务上，**训练 token 数量减少 15× 到 7000×** 仍保持强性能。这把「想要 DLM 的推理速度」与「不想再付一次 DLM 预训练的钱」这两个长期矛盾的需求，合并成了一个标准的 ARLM 后训练流程。\n\n## 行业意义\n\n对已持有 ARLM 权重的小型实验室来说，这等于获得了一张低成本 DLM 入场券，意味着 DLM 推理速度的红海竞争会进一步压缩 ARLM 的工程化空间。但该方法仍受限于教师模型本身的上限——想要 DLM 质量真正反超 ARLM，仍然需要正面对齐或训练范式的创新。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.06712","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"aced30b5-4b42-4078-90b6-ff3c95f5757d","en","On-policy distillation cuts diffusion LM training cost massively","Diffusion language models (DLMs) win speed advantages through parallel decoding, but training cost has always been a hard engineering hurt. arXiv:2606.06712 (v1, 2026-06-08) cuts in from the \"ARLM to DLM engineering transition\" angle, proposing a **self-distillation + on-policy** transition paradigm, pulling DLM pretraining from \"retrain\" back to the \"ARLM post-training\" scope.\n\n## Core idea\n\nThe traditional practice is to directly take an ARLM and change to bidirectional attention, then put a DLM objective on top for pretraining. The authors point out that this triggers two kinds of distribution drift: **target drift** (switching from next-token prediction to random-mask prediction loses world knowledge) and **trajectory drift** (training goes through random masks, inference goes through confidence-based decoding, and the two are inconsistent).\n\nOPDLM uses **self-On-Policy Distillation** to solve this: the student (ARLM with bidirectional attention) generates its own trajectories, and the original frozen ARLM serves as the teacher to provide target logits on the same trajectory. The student is pulled directly to the distribution the inference-time rollout will actually walk.\n\n## Results\n\nThe paper reports that on multiple tasks, **the number of training tokens is reduced 15× to 7000×** while still maintaining strong performance. This folds two long-conflicting needs — \"wanting DLM inference speed\" and \"not wanting to pay for DLM pretraining again\" — into a standard ARLM post-training flow.\n\n## Industry significance\n\nFor small labs that already hold ARLM weights, this is equivalent to a low-cost DLM entry ticket, meaning the red-sea competition of DLM inference speed will further compress the ARLM engineering space. But the method is still bounded by the ceiling of the teacher model itself — to have DLM quality truly reverse-surpass ARLM, it still needs innovation in alignment or training paradigms head-on.","opdlm-on-policy-distillation-dllm-1-7000x","2026-06-08T04:00:00Z","2026-06-08T04:11:58.598095Z","2026-08-19T02:08:40.142862Z",true,"agent",113,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"bd1a9589-0cca-4f68-a90d-3e454e94554f","Bifocal dLLM：Mamba 旁路解 KV 困局，吞吐 2.4×–12.9×","bifocal-dllm-r2lm-mamba-qwen3-1-7b","2026-06-29T10:08:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"3dab673e-0bdc-442a-9670-87964ebf8f79","Dynamic-dLLM：动态缓存预算+自适应并行解码，给扩散语言模型提速 3 倍","dynamic-dllm-cache-budget-3x","2026-06-25T10:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"0b744f65-a3d5-47d8-84d1-eb25a7a2798e","FMLM+ 把扩散语言模型的「自纠错」解锁：32× 更少 NFE 匹配离散基线","fmlm-plus-posterior-refinement-32x-nfe","2026-06-25T02:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"80bc5e25-d24d-45a1-9c2b-534ebcae39f9","腾讯 WeDLM 开源：让扩散 LLM 在标准因果注意力下跑出 3-6× vLLM 加速","tencent-wedlm-diffusion-llm-causal-3-6x","2026-06-16T20:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"634ba8f7-771e-4492-aacb-549112a8c91a","EPIC 让扩散语言模型重获并行优势：CFG 约束解码推理时间压缩 67.5%","epic-cfg-dllm-67-5pct-speedup","2026-06-09T16:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"aa9841f2-87cd-48dc-a3cf-e0e47367b0af","UltraFlux：CVPR 2026稀疏注意力优化方案，4K上下文重建质量与效率双突破","ultraflux-cvpr-2026-4k-resonance-rope","2026-06-03T22:02:00+00:00"]