[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-dspark-deepspec-confidence-scheduled":3,"topics-all":36,"news-related-625069f8-5163-4c70-88f8-10779cdf66e2":65},{"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},"625069f8-5163-4c70-88f8-10779cdf66e2","DSpark 把推测解码拉进「半自回归 + 置信度调度」时代:DeepSeek 联合北大在生产环境跑出 60%-85% 端到端提速","DSpark 把推测解码拉进「半自回归 + 置信度调度」时代:DeepSeek 联合北大在生产环境跑出 60%-85% 端到端提速\n\n6 月 27 日,DeepSeek 联合北京大学发布推测解码新框架 DSpark(Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation),创始人梁文锋位列作者。这不是又一次「主模型权重升级」,而是一次针对 LLM 在线推理系统瓶颈的工程级突围:在生产 V4-Flash\u002FV4-Pro 引擎上,相比已服役的 MTP-1 基线,DSpark 在相同吞吐下让单用户生成速度提升 60%-85%,并在 120 TPS 这类高交互 SLA 下把系统可用吞吐抬升 6.6 倍。\n\nDSpark 的解法分两层。草稿端,主体沿用并行 backbone,尾部接一个轻量 sequential block(Markov 头或 RNN 头),把 block 内的局部依赖补回来,既保住并行 drafter 的吞吐,又压住 suffix decay。验证端,confidence head 估计每位置的前缀存活概率,再叠加感知硬件队列的 scheduler,动态决定本次该验证多长,把 target 算力优先投向最可能接受的 token。\n\n实验层面,论文在 Qwen3-{4B,8B,14B} 与 Gemma4-12B 上对齐训练数据与草稿层数,把 DSpark 与 Eagle3(自回归)、DFlash(并行)同台比较:Qwen3-4B\u002F8B\u002F14B 上,macro-average accepted length 在 Eagle3 基础上提升 30.9%\u002F26.7%\u002F30.0%,在 DFlash 基础上提升 16.3%\u002F18.4%\u002F18.3%。三组基准(数学、代码、日常对话)整体上移,不是单一任务偏科。\n\n更值得关注的是生产部署:80 tok\u002Fs\u002Fuser 的中等 SLA 下,DSpark 让 V4-Flash 聚合吞吐提升 51%;120 tok\u002Fs\u002Fuser 的严格 SLA 下,基线 MTP-1 已逼近并发极限,DSpark 仍跑出 6.6 倍可用吞吐。换言之,DSpark 把「延迟-吞吐」这条 Pareto 边界整体外推,让过去「为保交互延迟只能压低并发」的高敏感场景第一次成为默认可选项。\n\n配套开源的 DeepSpec 是另一看点:把 DSpark、DFlash、Eagle3 三种 drafter 的训练\u002F评估代码统一到同一全栈仓库,MIT 协议,同步在 Hugging Face 释放了 Qwen3 与 Gemma4 系列的 12 个训练好的草稿 checkpoint——下游团队可以直接拿去域适配,这才是 DSpark 变成「行业基础设施」的关键一步。","https:\u002F\u002Fgithub.com\u002Fdeepseek-ai\u002FDeepSpec","4194681c-1a38-405d-a917-40e1dc2622ea",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"e02d6ab8-a980-4585-b9d5-98e468f25852","en","DSPark: semi-autoregressive speculation, 60-85% in production","On June 27, DeepSeek and Peking University released DSpark (Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation), with founder Liang Wenfeng listed as a co-author. This is not another \"main-model weight upgrade\" — it's an engineering-grade breakthrough targeting the online LLM inference system bottleneck: on the production V4-Flash \u002F V4-Pro engine, DSpark improves per-user generation speed by 60%–85% at the same throughput versus the in-service MTP-1 baseline, and at strict 120 TPS interactive SLAs it raises usable system throughput by 6.6×.\n\nDSpark's solution has two layers. On the drafter side, the body uses a parallel backbone, with a lightweight sequential block (Markov head or RNN head) at the tail to recover local dependencies inside the block — preserving the throughput of the parallel drafter while keeping suffix decay under control. On the verifier side, a confidence head estimates the per-position prefix-survival probability, combined with a hardware-queue-aware scheduler that dynamically decides how long to verify this round, directing target compute at the tokens most likely to be accepted.\n\nThe experiments align training data and drafter depth on Qwen3-{4B, 8B, 14B} and Gemma4-12B, and compare DSpark head-to-head with Eagle3 (autoregressive) and DFlash (parallel): on Qwen3-4B\u002F8B\u002F14B, the macro-average accepted length improves 30.9% \u002F 26.7% \u002F 30.0% over Eagle3 and 16.3% \u002F 18.4% \u002F 18.3% over DFlash. All three benchmark families (math, code, daily chat) move up, not just one task.\n\nMore important is production deployment: at a moderate 80 tok\u002Fs\u002Fuser SLA, DSpark lifts V4-Flash aggregate throughput by 51%; at a strict 120 tok\u002Fs\u002Fuser SLA where the MTP-1 baseline is already at the concurrency limit, DSpark still pulls 6.6× usable throughput. In other words, DSpark pushes the \"latency-throughput\" Pareto frontier outward as a whole, making the \"compress concurrency to preserve interactive latency\" high-sensitivity scenario a default option for the first time.\n\nThe companion open-source DeepSpec is another highlight: it unifies the training and evaluation code for DSpark, DFlash, and Eagle3 in one full-stack repo under MIT, with 12 trained drafter checkpoints for the Qwen3 and Gemma4 series released on Hugging Face. Downstream teams can take them for domain adaptation directly — and that's the key step for DSpark to become \"industry infrastructure.\"","dspark-deepspec-confidence-scheduled","2026-06-29T00:00:00Z","2026-06-29T00:06:50.364516Z","2026-08-19T02:08:40.142862Z",true,"agent",196,[37,47,56],{"slug":38,"tag_slug":39,"title_zh":40,"title_en":41,"intro_zh":42,"intro_en":43,"id":44,"is_active":33,"created_at":45,"modified_at":46},"image-generation","text-to-image","图像生成：从文生图到逐框编辑的模型竞赛","Image Generation: from text-to-image to frame-level editing","文生图已是红海，战场正转向编辑与可控性：FLUX.2\u002F3 用开源和量化把生图搬进消费级显卡，Qwen-Image、Seedream 等国产模型紧咬不放，GPT Image 2 则把推理能力带进像素。本专题追踪图像生成与编辑模型的发布、架构演进、加速量化与商业化数据，看「好图」如何变成「可用的图」。","Text-to-image is a crowded race, and the frontier has moved to editing and control: FLUX.2\u002F3 push open weights and quantization onto consumer GPUs, Qwen-Image and Seedream close in from China, and GPT Image 2 brings reasoning into pixels. This topic tracks model releases, architecture shifts, acceleration work, and the business numbers behind \"pretty\" becoming \"usable\".","52de7c59-7864-4569-8ae7-3d656d26a1a1","2026-11-20T07:20:23.534506Z","2026-11-20T07:20:23.534527Z",{"slug":48,"tag_slug":48,"title_zh":49,"title_en":50,"intro_zh":51,"intro_en":52,"id":53,"is_active":33,"created_at":54,"modified_at":55},"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":57,"tag_slug":57,"title_zh":58,"title_en":59,"intro_zh":60,"intro_en":61,"id":62,"is_active":33,"created_at":63,"modified_at":64},"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":66},[67,72,77,82,87,92],{"id":68,"title":69,"news_slug":70,"published_at":71},"6a099c03-4df1-434f-acbf-4a6a675f9ce7","神经草稿变慢变耗电:0.2毫秒查表草稿换来2.43倍","olifant-memory-draft-speculative-decoding","2026-11-14T09:20:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"9cfac3fd-f359-467e-94ba-f8ed3172d713","生成器自己当重排器:RAG 少驻留一个模型","glance-generator-reranker-shared-prefill","2026-10-31T13:11:54+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"559df680-f13e-48f0-9263-73116aaff411","上交大DLite免历史投机解码:Agent提速2.74倍","dlite-history-free-speculative-decoding","2026-10-19T17:07:52+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"2638aeac-dc4d-4b73-b7fe-2b042015adee","OreoLook 开源:三层缓存把 AI 搜索搬进 8 核 CPU,重复问题 0.1 毫秒出答案","oreolook-three-layer-cpu-cache","2026-09-10T23:08:36+00:00",{"id":88,"title":89,"news_slug":90,"published_at":91},"0fe9ceb8-6411-4924-8e02-8cee3665fc6f","Cohere 开源 megakernel 推理引擎：单 CUDA 文件，H100 解码吃到 62% 带宽光速","cohere-megakernel-north-mini-code-h100","2026-09-08T21:13:46+00:00",{"id":93,"title":94,"news_slug":95,"published_at":96},"0c29e1ad-914a-4b79-a153-445c087acb03","被 LLM 抛弃的 dropout 翻身:Cerebras 称调好可省 25% 训练 FLOPs","dont-drop-dropout-layer-sparsity","2026-09-07T21:06:35+00:00"]