[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-solarwm-open-data-video-world-models":3,"topics-all":38,"news-related-fb2da954-b12e-4bde-9146-b61dd240df92":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},"fb2da954-b12e-4bde-9146-b61dd240df92","SolarWM 开源:143 万条视频喂出的世界模型,5 秒训练片段撑起小时级交互","港中深为首九机构团队开源交互式视频世界模型全栈 SolarWM:143 万条剪辑统一成帧对齐数据契约,一套三阶段配方把 Wan2.2、LTX-2.5、MiniMax-H3 三种骨干改造成实时交互模型,只用 5 秒训练片段即可推出分钟到小时级 rollout,代码数据权重全部放出。","世界模型的演示视频这两年看了不少,但真想复现一个,地狱全在细节里:各家数据集的时间尺度、相机几何、字幕风格互不兼容,训练配方又和单一骨干绑死,论文之间几乎没法公平比较。9 月 3 日,一个以港中大(深圳)为首、联合 NVIDIA、MSRA 等九家机构的团队开源了 SolarWM,试图把这件事一次性捋直。\n\n## 数据层:143 万条剪辑,先统一成一份契约\n\nSolarWM 的第一步不是训模型,是修数据。它把 14 个数据集的 143 万条规范剪辑转换成统一的帧对齐契约,覆盖视觉观测、度量级相机几何、字幕、质量元数据、选择决策和来源信息,并且把源数据处理和训练混合设计解耦——想换数据配比,不用重跑整条管线。配套的 SolarWM-Data 已经上架 Hugging Face 和 ModelScope。\n\n## 一套配方,三种骨干\n\n框架在同一套相机条件化、训练和推理接口之下,实例化了四个 5B 到 33B 的模型,底座分别是 Wan2.2、LTX-2.5 和 MiniMax-H3,且各自保留原生表示与训练目标。训练路线是三段式:Stage0.5 做全片段双向 flow matching,建立视频、文本、相机条件的基础表示;Stage1 用教师强制加 AnyFlow 损失,让干净历史去条件化带噪目标块,同时学去噪和有限步流映射——论文强调这省掉了 DMD 之前单独的 ODE 或一致性蒸馏初始化;Stage2 用自梯度 forcing 做分布匹配蒸馏,因果学生模型在自己的 rollout 上训练,教师冻结、critic 可训。\n\n## 5 秒片段,小时级交互\n\n最有说服力的结果是:全部训练只用了 5 秒序列,得到的因果模型却能实时交互,rollout 长度从分钟级到小时级,不需要长序列微调,也不依赖 attention-sink 机制。项目代码、数据管线、配方和权重全部放出,代码走 Apache 2.0,GitHub 上线首日 249 star。\n\n## 泼点冷水\n\n细看 README,有三处值得冷静。其一,阶段表里只有 Wan2.2-5B 走完了全部三个阶段,Wan2.2-14B、LTX-2.5、MiniMax-H3 三个底座目前只放出 Stage0.5 双向权重,Stage1 和 Stage2 都标着 Coming soon。其二,公开数据发布并不包含完整视频载荷:latent 版按配方下载即可训练,raw-WDS 要么用 annotation 包从原始发布方自行重建,要么填表申请。其三,开源也有小字:代码是 Apache 2.0,但 LTX 衍生权重遵循 LTX 社区许可,MiniMax H3 许可含地域限制条款。\n\n## 所以呢\n\n从 ForgeWM、EchoWM 到 SolarWM,两个月内世界模型赛道连续出现全流程开源,方向很清楚:竞争正从谁的视频更好看,转向谁的数据契约和训练配方更可复用。SolarWM 这份跨骨干配方能不能被社区接住,就看那三格 Coming soon 什么时候填满。\n\n参考:arXiv:2609.02886;github.com\u002FJunchao-cs\u002FSolarWM\n","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.02886","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},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":19,"name":20,"slug":20,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",{"id":22,"name":23,"slug":23,"description":14,"color":14},"ebe5dcd1-46b1-4298-b8c2-8e0e2f456e56","video-generation",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"f96f262b-2cc4-496f-b095-dcc351460920","en","SolarWM: Open Data and Recipes for Video World Models","SolarWM open-sources a world-model stack: 1.43M clips in one data contract, one 3-stage recipe across three backbones, hour-long real-time rollouts.","World-model demos are everywhere, but reproducing one is painful: datasets differ in temporal scale, camera geometry and captioning style, while training recipes stay locked to a single backbone. On September 3, a team led by CUHK-Shenzhen together with NVIDIA, MSRA and six other institutions open-sourced SolarWM, an attempt to straighten all of this out in one release.\n\n## The data layer: 1.43M clips under one contract\n\nSolarWM starts with data, not models. It converts 1.43 million canonical clips from 14 datasets into a unified, frame-aligned contract covering visual observations, metric camera geometry, captions, quality metadata, selection decisions and provenance. Source processing is decoupled from training-mixture design, so changing the mixture does not require rerunning the pipeline. The SolarWM-Data release is live on Hugging Face and ModelScope.\n\n## One recipe, three backbones\n\nUnder shared camera-conditioning, training and inference interfaces, the framework instantiates four 5B-33B models built on Wan2.2, LTX-2.5 and MiniMax-H3, each preserving its native representation and objective. The recipe has three stages: Stage0.5 learns full-clip bidirectional flow matching to establish base video, text and camera-conditioned representations. Stage1 combines teacher forcing with the AnyFlow loss, conditioning noisy target chunks on clean history while learning both denoising and finite-step flow maps, which the paper says removes the need for a separate ODE or consistency-distillation initialization before distillation. Stage2 runs distribution matching distillation via self-gradient forcing, training the causal student on its own rollout with a frozen teacher and a trainable critic.\n\n## 5-second clips, hour-long interaction\n\nThe headline result: training used only 5-second sequences, yet the resulting causal models support real-time interaction with rollouts spanning minutes to hours, without long-sequence fine-tuning or attention-sink mechanisms. Code, data pipeline, recipes and weights are all out, with the code under Apache 2.0 and 249 GitHub stars on day one.\n\n## Some cold water\n\nThree caveats in the README. First, only Wan2.2-5B has completed all three stages; the other three backbones currently ship Stage0.5 bidirectional weights only, with Stage1 and Stage2 marked Coming soon. Second, the public data release omits full raw video payloads: latent versions download per recipe, while raw-WDS must be rebuilt from the annotation package or requested via an access form. Third, openness has fine print: the code is Apache 2.0, but LTX derivatives follow the LTX community license and the MiniMax H3 license carries territory restrictions.\n\n## So what\n\nFrom ForgeWM and EchoWM to SolarWM, the world-model lane has produced three full-stack open releases in two months. The competition is shifting from whose video looks better to whose data contracts and training recipes are reproducible. Whether the community picks up this cross-backbone recipe depends on when those three Coming-soon cells get filled.\n\nReferences: arXiv:2609.02886; github.com\u002FJunchao-cs\u002FSolarWM\n","solarwm-open-data-video-world-models","2026-09-03T15:08:13Z","2026-09-03T15:09:25.278467Z","2026-09-03T15:09:25.278478Z",true,"agent",224,[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},"2874a2e5-beae-4627-8f6f-a34cf2cc8d7a","一段随手拍视频直出4D人体:4DAnyone用RCP+TCR破解多视角一致性,代码权重全开源","4danyone-monocular-video-4d-human","2026-08-20T17:59:53+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"6f9e9f94-9dcc-4c6c-b254-6c5d0fe8ed37","京东开源 JoyAI-Video-Edit:16B 多模态扩散 Transformer 把视频编辑推进「边播边改」实时流时代","jd-joyai-video-edit-realtime-diffusion","2026-08-10T00:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"bcdc10bc-2f08-4c39-8ffa-e7e34041c112","京东开源 JoyAI-Video-Edit:用 16B 多模态扩散 Transformer 把视频编辑推进「边播边改」实时流时代","jd-joyai-video-edit-real-time-streaming","2026-08-05T03:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"e840fad5-b3a1-47cd-ac68-679d5f635dc1","世界状态交给程序管:Programmable World Model 让视频模型只管渲染","programmable-world-model-persistent-state","2026-09-10T17:10:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"caed836e-2168-418f-b5c1-bde3ce962e66","Mask Forcing 往蒸馏 rollout 里掺干净 token:修视频生成的模式坍缩,指令遵循最高涨 6.5 分","mask-forcing-video-diffusion-distillation","2026-09-09T23:08:37+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"7ddc323f-fc52-406a-b6df-79b7393e121b","高德开源 DreamX-Creator:7B 原生音视频生成,2K 输出","dreamx-creator-7b-native-audio-video","2026-09-01T13:10:00+00:00"]