[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-programmable-world-model-persistent-state":3,"topics-all":38,"news-related-e840fad5-b3a1-47cd-ac68-679d5f635dc1":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},"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","Alaya Lab 提出 Programmable World Model:Agent 把自然语言编译成程序,轻量引擎维护持久世界状态,视频模型只当渲染器。CombatStateBench 计数准确率 94%、状态准确率 98%,项目已开源。","视频世界模型这两年进化飞快:画面越来越真,交互越来越顺。但有个问题一直没解决——玩得越久,世界越\"糊涂\"。敌人还剩几个?道具用掉没?角色走出画面后状态还在不在?这些信息只存在于生成帧里,模型全靠\"感觉\"续写,长交互一多就露馅。9 月 9 日提交到 arXiv 的论文 [Programmable World Model](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.10540)(Alaya Lab,11 位作者)给出了一条干脆的路线:别让画面管状态,状态交给程序管。\n\n## 问题:状态藏在帧里,迟早会漂\n\n论文摘要对现有视频世界模型的批评很直接:它们\"缺乏在长时间交互下维持持久世界状态、执行可编程规则的可靠机制\"。生成模型擅长的是\"下一帧长什么样\",而战斗计数、实体生死、规则约束这些是离散逻辑状态——让一个概率采样器去硬记,本质上是用错了工具。这也解释了为什么很多演示只能剪成十几秒的片段:拉长到分钟级,状态漂移就藏不住了。\n\n## 方案:程序管状态,视频只管渲染\n\n框架的核心是一次彻底解耦,分三层:\n\n- **Agent 写程序**:把自然语言指令翻译成可执行程序,程序里定义实体状态和状态转移规则,可以直接控制单个实体及其交互;\n- **轻量引擎维护状态**:执行这些程序,更新并维护一个显式、持久的全局世界状态——注意,包括画面外的实体和非视觉属性,这是纯视频路线天然做不到的;\n- **视频模型当渲染器**:状态与画面之间用带状态的 3D 有向包围盒(OBB)做中间表示,连同目标相机轨迹,确定性编译成像素级对齐的时空条件信号,喂给一个预训练视频模型做\"生成式渲染\"。\n\n这套设计直接支持带预设机制的可玩游戏:用户能单独操控某个实体,游戏全程状态在线不丢。\n\n## 结果:CombatStateBench 状态准确率 98%\n\n团队同步发布了评测基准 CombatStateBench,专测可编程世界模型。论文报告:计数准确率 94%、状态准确率 98%,并支持连贯的长时域生成,优于现有交互式视频世界模型——这是作者自报口径,具体名次等独立复现,但\"状态用程序维护\"这个差异点本身已经足够清晰。项目已在 [GitHub 开源(AlayaLab\u002Fpwm)](https:\u002F\u002Fgithub.com\u002FAlayaLab\u002Fpwm),发布当天冲上 Hugging Face Daily Papers 热榜前排。\n\n## 评论:把游戏引擎最成熟的部分还给世界模型\n\n这件事的本质,是把经典游戏引擎里最可靠的东西——显式状态机——重新塞回生成式世界模型。纯端到端路线赌\"模型足够大就会自己学会记状态\",这条路线则承认:强逻辑、需要审计的状态,程序比概率可靠。对做交互内容的人来说,可编程意味着可控、可调试、可复现,而这恰恰是生成模型最缺的三样。值得盯的下一步是:这套\"程序+渲染\"的分工,能不能撑住比战斗场景更开放的世界——如果能,世界模型的落地路径会短很多。\n","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.10540","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",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},"09934abd-2a73-4daa-8aa4-8c93897850be","en","Programmable World Model: Programs Manage State, Video Only Renders","Alaya Lab's Programmable World Model decouples state from rendering: an agent compiles natural language into programs, a lightweight engine maintains persistent world state, and a pretrained video model serves as renderer. Reports 94% count accuracy and 98% state accuracy on CombatStateBench; project open-sourced.","Video world models have advanced rapidly: visuals grow more realistic and interaction smoother. But one problem persists—the longer you play, the hazier the world gets. How many enemies remain? Was the item consumed? Does a character's state survive after walking off-screen? This information lives only inside generated frames, and the model keeps going by feel; extended interactions expose the cracks. A paper submitted to arXiv on September 9, [Programmable World Model](https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.10540) (Alaya Lab, 11 authors), offers a decisive route: stop letting the pixels manage state—let programs manage it.\n\n## The Problem: State Hidden in Frames Drifts\n\nThe paper's abstract is blunt about existing video world models: they \"lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions.\" Generative models excel at \"what the next frame looks like,\" while combat counts, entity life-and-death, and rule constraints are discrete logical state—forcing a probabilistic sampler to memorize them is simply using the wrong tool. This also explains why many demos are cut into ten-second clips: stretch to minutes and state drift becomes impossible to hide.\n\n## The Approach: Programs Manage State, Video Only Renders\n\nThe framework's core is a thorough decoupling in three layers:\n\n- **The agent writes programs**: natural-language instructions are translated into executable programs that specify entity states and state-transition rules, enabling direct control over individual entities and their interactions;\n- **A lightweight engine maintains state**: it executes these programs to update and maintain an explicit, persistent global world state—including off-screen entities and non-visual attributes, something a pure video approach fundamentally cannot do;\n- **A pretrained video model serves as renderer**: state connects to visuals via state-augmented 3D oriented bounding boxes (OBBs) as an intermediate representation, which, together with the target camera trajectory, is deterministically compiled into pixel-aligned spatiotemporal conditioning signals fed to the generative renderer.\n\nThis design directly supports playable games with predefined mechanics: users can control individual entities directly, with persistent world state throughout gameplay.\n\n## Results: 98% State Accuracy on CombatStateBench\n\nThe team also released CombatStateBench, a benchmark for evaluating programmable world models. The paper reports 94% Count Accuracy and 98% State Accuracy, outperforming existing interactive video world models while supporting coherent long-horizon generation—figures are self-reported by the authors pending independent replication, but the differentiation of \"state maintained by programs\" is already clear enough. The project is [open-sourced on GitHub (AlayaLab\u002Fpwm)](https:\u002F\u002Fgithub.com\u002FAlayaLab\u002Fpwm) and reached the front ranks of Hugging Face Daily Papers on its release day.\n\n## Commentary: Give World Models Back the Most Reliable Part of Game Engines\n\nThe essence of this work is re-inserting the most reliable component of classic game engines—the explicit state machine—into generative world models. The pure end-to-end route bets that \"a big enough model will learn to remember state on its own\"; this route concedes instead that strongly logical, auditable state is more reliably handled by programs than probabilities. For interactive-content creators, programmable means controllable, debuggable, and reproducible—precisely the three things generative models lack most. The next milestone to watch: whether this division of labor between \"programs + rendering\" can hold up in worlds more open than combat scenarios—if it can, the path to deployment for world models gets much shorter.\n","programmable-world-model-persistent-state","2026-09-10T17:10:00Z","2026-09-10T17:10:01.153425Z","2026-09-10T17:10:01.153439Z",true,"agent",99,[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},"dc2f4ead-963c-4a8e-bd41-400bebf83bb4","物理、几何、外观一个模型全包:Puffin-World 开源,相机 roll 误差低至 0.26°","puffin-world-native-3d-world-states","2026-09-06T19:09:41+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"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","solarwm-open-data-video-world-models","2026-09-03T15:08:13+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"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",{"id":75,"title":76,"news_slug":77,"published_at":78},"630ed9ae-699e-4115-8e75-33196ea6db28","MiniMax Music 3 开源:8B+0.6B 双 LLM 写五分钟完整歌,8GB 显存能跑","minimax-music3-open-weights-architecture","2026-08-29T13:30:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"b1400260-ba9f-4e84-b658-ce53abba9304","BreezeBlue 开源 Breeze TTS 2:3B 参数实时语音,五语种、可控设计、首包 133 毫秒","breeze-tts-2-open-source-realtime","2026-08-29T10:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"e3c0b314-d7b7-4901-b2b0-08ca5ef08ac7","GigaBrain-0.7开源:37k小时数据+三系统架构,世界模型进VLA决策回路","gigabrain-0-7-embodied-vla-open-source","2026-08-26T23:15:00+00:00"]