[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-uno-diffusion-augmented-llm-speedup":3,"topics-all":38,"news-related-8914cfad-53bc-4f76-bb06-8579d23cfa5d":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"8914cfad-53bc-4f76-bb06-8579d23cfa5d","AR 大模型外挂 diffusion 权重:Uno 免草稿模型实现 3 倍无损解码加速","arXiv 新论文 Uno 把轻量扩散权重嫁接到自回归 LLM 上,配合 Ψ-Spec 采样器并行抽 token,免草稿模型实现最高 3 倍无损加速;论文自报 8B 版在工具调用、编码与长上下文基准上超过 26B DiffusionGemma。","自回归 LLM 的命门是「逐 token 生成」:下一个词依赖上一个词,解码只能串行,GPU 再强也只能干等。投机解码用小草稿模型先猜再验,但草稿模型本身要训练、要显存,batch 一大加速比就垮。9 月 3 日挂上 arXiv 的论文《Unlocking Lossless Speedups in LLMs via Discrete Diffusion》(arXiv:2609.04010)给出第三条路:不动 AR 模型本体,给它外挂一组轻量 diffusion 权重,让多个 token 并行「抽」出来。作者来自 Institute of Foundation Models(IFM),合作方包括 Cerebras Systems 及多所高校。\n\n## 两套权重,各司其职\n\n论文提出 diffusion-augmented LLM 这一类新模型,把参数拆成两套:AR 权重照常用 next-token prediction 目标训练,负责质量;轻量 diffusion 权重负责一次并行生成多个 token。后者通过一个 Diffusion Distillation 阶段学出来,作者称对现有 LLM 训练管线的额外开销可以忽略。配套的 Ψ-Spec 采样器族从 AR 分布里并行采样,做到无损加速——输出仍服从原 AR 模型定义的分布,还支持在固定上下文长度下做推理时扩展。与投机解码最大的区别:不需要单独的草稿模型;与 diffusion LLM(d-LLM)的区别:不牺牲底层 AR 模型的质量。项目页对比图还显示,在附加参数量和推理显存上,Uno 都是受测方法里最低的一档。\n\n## 8B 打 26B,Pareto 压制投机解码\n\n实测结果两条。其一,Uno 在所有受测 batch size 下吞吐都高于主流投机解码方法(论文点名 DFlash 和 Eagle3),相对基础 AR 模型最高 3 倍加速——即便在设备支持的最大 batch 下也成立,项目页称之为对投机解码的 Pareto 支配。其二,论文自报 8B 的 Uno 在 agentic 工具调用、编码、长上下文推理的全部受测基准上,超过 26B 的 DiffusionGemma(论文称之为领先的开源 d-LLM)和闭源的 Mercury 2。这组数字目前是作者自测,尚待社区复现,但方向值得注意:加速不必再拿质量去换。\n\n## 所以呢\n\nUno 可以从零训练,也可以直接给现有开源权重的 AR 模型「加装」,代码与 checkpoint 已开源(github.com\u002Fifm-ai\u002Funo)。这篇论文 9 月 8 日登上 HuggingFace Daily Papers 日榜首位。对推理 infra 团队的启示:diffusion 不一定要替换 AR,它可以作为并行解码的外挂件嫁接进现有技术栈——如果你在生产环境跑投机解码,这条免草稿模型的路线值得拉一遍 benchmark 对照。论文原文:arxiv.org\u002Fabs\u002F2609.04010。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.04010","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":19,"name":20,"slug":20,"description":14,"color":14},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"f5c0bb81-789b-42f2-be94-4429d3781af6","en","Uno Bolt-On Diffusion Weights Give AR LLMs 3x Lossless Speedup","arXiv paper Uno adds lightweight diffusion weights to autoregressive LLMs, decoding tokens in parallel: up to 3x lossless speedup, no draft model needed.","Autoregressive LLMs have one built-in bottleneck: the next token depends on the previous one, so decoding is strictly sequential and GPUs spend much of their time waiting. Speculative decoding attacks this with a small draft model that guesses ahead, but the draft model itself needs training and memory, and its speedup collapses at large batch sizes. A paper posted to arXiv on September 3, \"Unlocking Lossless Speedups in LLMs via Discrete Diffusion\" (arXiv:2609.04010), proposes a third path: keep the AR model intact and bolt on a set of lightweight diffusion weights that draw multiple tokens in parallel. The work comes from the Institute of Foundation Models (IFM), with collaborators from Cerebras Systems and several universities.\n\n## Two weight sets, separate jobs\n\nThe paper introduces diffusion-augmented LLMs, a class of models with two parameter sets: AR weights trained with the standard next-token-prediction objective, responsible for quality, and lightweight diffusion weights trained to generate multiple tokens simultaneously. The latter are learned through a Diffusion Distillation phase that the authors say adds negligible overhead to existing LLM training pipelines. A family of samplers called Ψ-Spec draws tokens in parallel from the AR-defined distribution, enabling lossless acceleration and inference-time scaling at a fixed context length. Versus speculative decoding, the biggest difference is that no separate draft model is required; versus diffusion LLMs (d-LLMs), the quality of the underlying AR model is not sacrificed. The project page comparison also shows Uno introduces the least additional parameters and the lowest inference-time GPU memory among the evaluated methods.\n\n## An 8B model beating a 26B d-LLM\n\nTwo result lines stand out. First, Uno achieves higher throughput than leading speculative-decoding methods — the paper names DFlash and Eagle3 — at every evaluated batch size, with up to 3x speedups over the base AR model, including at the largest batch size supported by the device; the project page describes this as Pareto-dominating speculative decoding. Second, the authors report that the 8B Uno model outperforms the 26B DiffusionGemma, which the paper calls the leading open d-LLM, and the proprietary Mercury 2 across all evaluated benchmarks in agentic tool use, coding, and long-context reasoning. These numbers are self-reported and await community replication, but the direction matters: speedup without trading away quality.\n\n## So what\n\nUno can be trained from scratch or built by augmenting existing open-weight AR LLMs, and code plus checkpoints are already open (github.com\u002Fifm-ai\u002Funo). The paper also sat at the top of HuggingFace Daily Papers for September 8. For inference infra teams the lesson is that diffusion does not have to replace autoregression — it can be attached to an existing stack as a parallel-decoding bolt-on. If you run speculative decoding in production, this draft-free route is worth a benchmark pass. Full paper: arxiv.org\u002Fabs\u002F2609.04010.","uno-diffusion-augmented-llm-speedup","2026-09-08T13:30:00Z","2026-09-08T13:12:54.555501Z","2026-09-08T13:12:54.555509Z",true,"agent",168,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"created_at":55,"modified_at":56},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"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":65,"title":66,"news_slug":67,"published_at":68},"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":70,"title":71,"news_slug":72,"published_at":73},"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":75,"title":76,"news_slug":77,"published_at":78},"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":80,"title":81,"news_slug":82,"published_at":83},"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":85,"title":86,"news_slug":87,"published_at":88},"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","opdlm-on-policy-distillation-dllm-1-7000x","2026-06-08T04:00:00+00:00"]