[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-youtu-parsing-omni-5b":3,"topics-all":38,"news-related-faf5c85e-ed83-432a-8b55-80c9356862c5":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},"faf5c85e-ed83-432a-8b55-80c9356862c5","腾讯优图开源5B全模态解析:文档榜96.96超GPT-5.2","腾讯优图实验室开源 5B 全模态解析模型 Youtu-Parsing-Omni,单一编码器统一解析文档、图表、几何、音频与视频为结构化 JSON,OmniDocBench v1.6 拿下 96.96 居首,超过 Gemini-3-Pro 与 GPT-5.2,权重与 vLLM 插件已放出。","文档解析大概是多模态 AI 里最不性感的赛道——没有对话,没有生成,就是把一页纸变成机器能用的结构化数据。但腾讯优图实验室新开源的 Youtu-Parsing-Omni 交出了不太寻常的成绩单:5B 参数,在 OmniDocBench v1.6 文档解析基准上拿到 96.96 的总分,身后是 Gemini-3-Pro(92.91)、Qwen3-VL-235B(89.78)和 GPT-5.2(86.59)。\n\n## 一个模型,十类任务\n\n它做的事一句话就能说清:给它文档页、自然图像、图表、几何图形、音频或一段带声音的视频,它输出统一的结构化 JSON——感知层给版面元素、文本、表格、公式、边界框、时间戳、ASR、声学事件,认知层给标题、叙述、报告,十类任务共用一套 OmniSchema 输出格式。\n\n架构上它没走常见的「双塔」路线(视觉、音频各一座塔,到语言模型里才汇合),而是用单一的 Youtu-Omni-Encoder 吃下所有模态;编码器从预训练文本语言模型初始化以强化文字读取,像素和 log-mel 音频帧经薄 stem 转成 token,由共享双向 Transformer 在统一的 (t, h, w) 位置编码下做上下文化——视觉与音频共享同一时间坐标系,跨模态对齐在编码器内部完成,而非留给解码器事后重建。\n\n## 后训练:让学生自己当老师\n\n训练路径是 SFT 加 schema 路由的 RLVR,再加一段名为 OSAD 的自演化后训练:学生采样自己的解析,通过准入门的 rollout 由一个能看到参考解析的冻结副本打分;每轮结束合并 LoRA 适配器、学生变成下一轮的老师——不需要手工设计的奖励,也不需要单独训练教师模型。\n\n## 跑分拆开看:垂直碾压,通用还差口气\n\n文档侧 OmniDocBench v1.6 上,96.96 总分、98.20 表格 TEDS-S、0.1104 阅读顺序编辑距离都是第一;化学式识别 ChemOCR 上 Tani 相似度 54.2 居首,GPT-4o 只有 3.4;乐谱识别 PDMX-Synth 上字符错误率 22.73,优于专门的乐谱模型 LEGATO(23.30),GPT-5.6 是 45.02。\n\n但全模态侧的 OmniParsingBench 讲了另一个故事:均值 75.08 排第二,落后 Gemini-3-Pro 的 77.44。细分看,图表(95.02)和音频(78.08)两项第一,几何图形(74.33 对 85.43)、自然图像、自然视频明显落后。解析任务上已能打旗舰,通用视觉理解还有差距——官方没回避,表就在 model card 里。\n\n## 所以呢\n\n权重已在 Hugging Face 和 ModelScope 放出,配套 vLLM、SGLang 插件与 Docker 镜像,技术报告 PDF 放在 GitHub 仓库(TencentCloudADP\u002Fyoutu-parsing),arXiv 编号待补;许可证为自定义条款,商用前值得读一遍。\n\n比模型更值得注意的信号是:解析类脏活正从「调用旗舰 API」转向「专用小模型本地跑」。5B 一台工作站就能部署,对手是按 token 计费的 GPT-5.2 和 Gemini-3-Pro。对知识库、票据流水线、音视频归档团队,这不是又一个「开源追平闭源」的故事,而是成本结构换了一档。当脏活被 5B 模型承包,旗舰 API 的护城河只剩通用推理。\n\n参考:huggingface.co\u002Ftencent\u002FYoutu-Parsing-Omni(model card 与评测表)、github.com\u002FTencentCloudADP\u002Fyoutu-parsing(技术报告 PDF 与推理代码)。","https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FYoutu-Parsing-Omni","b6764422-ee82-4f78-b924-1d43742e4669",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"1a3d4e7f-84aa-49d7-a24b-cf63daeaa3b1","en","Tencent's Youtu-Parsing-Omni: a 5B Model Out-Parsing GPT-5.2","Tencent open-sources Youtu-Parsing-Omni, a 5B model parsing documents, audio and video into structured JSON; 96.96 on OmniDocBench tops GPT-5.2.","Document parsing is arguably the least glamorous corner of multimodal AI — no chat, no generation, just turning a page into machine-usable structured data. Yet Youtu-Parsing-Omni, the model Tencent's Youtu Lab just open-sourced, posted a rather unusual scorecard there: 5B parameters, 96.96 overall on the OmniDocBench v1.6 document parsing benchmark, with Gemini-3-Pro (92.91), Qwen3-VL-235B (89.78) and GPT-5.2 (86.59) lined up behind it.\n\n## One Model, Ten Task Types\n\nWhat it does can be stated in one sentence: feed it a document page, a natural image, a chart, a flowchart, a geometry figure, an audio clip or an audio-visual video, and it emits one structured JSON envelope — covering both perception (layout elements, text, tables, formulas, bounding boxes, timestamps, ASR, acoustic events) and cognition (captions, narratives, reports). Ten parsing tasks share one output schema, OmniSchema.\n\nArchitecturally it skips the common two-tower design where vision and audio each get a separately pre-trained encoder that only meet inside the language model. Youtu-Parsing-Omni replaces both towers with a single Youtu-Omni-Encoder initialized from a pre-trained text language model, which naturally strengthens text reading. Pixels and log-mel audio frames pass through thin modality-specific stems into tokens; a shared bidirectional Transformer contextualizes an arbitrary packed mix of images, audio chunks and videos under one (t, h, w) positional encoding — vision tokens take their time coordinate from the frame index and spatial ones from the patch grid, audio tokens take time from their offset with spatial coordinates set to zero, so both modalities share a single temporal coordinate system. Cross-modal alignment is thus computed inside the encoder rather than reconstructed by the decoder; a few fusion layers open a new attention window at every audio-to-frame boundary so each frame attends to the sound of the same instant.\n\n## Post-Training: the Student Becomes the Teacher\n\nThe training path is SFT plus schema-routed RLVR, followed by a self-evolving post-training procedure called OSAD (OmniSchema-Aware On-Policy Distillation): the student samples its own parses; rollouts that pass an admission gate (JSON validity, schema conformance, box syntax, repetition checks) are scored by a frozen copy of the model that additionally sees the reference parse. Content tokens get a Jensen-Shannon divergence objective, structure tokens a forward KL at higher temperature, and at every round boundary the student's LoRA adapters are merged and the student becomes the next teacher — no hand-designed reward, no separately trained teacher model.\n\n## Reading the Benchmarks: Vertical Domination, General Vision Still Behind\n\nThe two most telling tables in the official evaluation are these. On the document side, OmniDocBench v1.6 gives it 96.96 overall, 98.20 table TEDS-S and a 0.1104 reading-order edit distance, all first; on ChemOCR chemistry recognition its Tani similarity of 54.2 leads the board while GPT-4o sits at 3.4; on the PDMX-Synth sheet-music benchmark its character error rate of 22.73 beats the specialized OMR model LEGATO (23.30), with GPT-5.6 at 45.02.\n\nBut the omni-modal OmniParsingBench tells another story: 75.08 average, second place, behind Gemini-3-Pro's 77.44. Broken down, it leads on charts (95.02) and audio (78.08) but trails clearly on geometry (74.33 vs 85.43), natural images and natural video. In other words, on its home turf of parsing tasks it already fights flagships; on general visual understanding there is still a gap — a gap the team did not hide, since the table sits right in the model card.\n\n## So What\n\nWeights are up on Hugging Face and ModelScope, with vLLM and SGLang plugins plus Docker images; the technical report ships as a PDF in the GitHub repo (TencentCloudADP\u002Fyoutu-parsing) with the arXiv ID still pending; the license is a custom non-standard template worth reading before commercial use.\n\nThe signal worth watching is bigger than the model itself: parsing chores are migrating from \"call a flagship API\" to \"run a specialized small model locally\". A 5B footprint means one workstation can host it, while the comparison targets are token-billed GPT-5.2 and Gemini-3-Pro. For teams building knowledge bases, invoice pipelines or audio-visual archives, this is not another \"open-source catches up\" story — the cost structure simply drops a tier. When a 5B model takes over the dirty work, the flagship APIs' moat shrinks to general reasoning; that part, for now, nobody can replace.\n\nReferences: huggingface.co\u002Ftencent\u002FYoutu-Parsing-Omni (model card and eval tables), github.com\u002FTencentCloudADP\u002Fyoutu-parsing (technical report PDF and inference code).","youtu-parsing-omni-5b","2026-10-19T19:09:27Z","2026-10-19T19:10:27.241656Z","2026-10-19T19:10:27.241663Z",true,"agent",144,[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},"b44fb160-9027-44cd-b355-a66869004abc","Nemotron 3.5 Super VL 开源:120B 总参只激活 12B","nemotron-3-5-super-vl-open-120b-moe","2026-10-20T17:11:56+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"ac666c4e-4756-43d8-840a-4eacbf88f804","Meta 开源 816B MoE 模型 Muse Spark Open,跑在两台 Mac Studio 上","meta-muse-spark-open-816b-mac-studio","2026-10-17T03:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"1b6ec126-cba1-47da-870b-da19cc523763","Meta 开源 816B MoE 模型 Muse Spark Open:1M 上下文,两台 Mac Studio 跑得动","meta-muse-spark-open-816b-moe-1m-mac-studio","2026-10-15T11:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"5d4acf64-67d8-404b-9d3f-cdf7761fb8ac","小鹏开源 GroundingPI:4B 视觉 grounding 模型自评超 GPT-6 Astra","groundingpi-4b-visual-grounding-foundation-model","2026-10-13T17:10:26+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"cd49f913-cde7-4cf3-8d93-24508653180e","腾讯混元开源AuK:1.5B语音模型统一生成与编辑,4步推理快4.5倍","tencent-hunyuan-auk-speech-editing","2026-09-09T09:12:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"40095b51-97b0-4fd4-9b1d-f636c970572e","阿里 Qwen 团队发布 Qwen3.8-Max:2.4 万亿参数 MoE 模型首度开放权重","qwen3-8-max-2-4t-moe-open-weights","2026-08-07T02:00:00+00:00"]