[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-jepa-anything-orthogonal-predictive-factorization":3,"topics-all":38,"news-related-12c67d52-17a2-4df5-8386-35d18ffd221a":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},"12c67d52-17a2-4df5-8386-35d18ffd221a","JEPA-Anything:一套预测框架打通七个领域,湿实验也给了背书","arXiv 2609.20800 提出 JEPA-Anything:用正交预测分解让七个领域(视觉、生物、临床、控制、分子、物理场、天气)共享一套预测原理。10 个动力学任务全面提升,干预误差降 34.8%,四系统分子 rollout 最低,开普勒指数拟合 -1.4991,生物干预获湿实验支持。","大模型圈这两年把「世界模型」喊得很响,但绝大多数预测模型其实各自为战:视觉一套架构、分子动力学一套架构、临床轨迹又是一套。9 月 17 日挂在 arXiv 上的 JEPA-Anything(编号 2609.20800)试图回答一个更根本的问题——细胞、分子、物理场、天气这些截然不同的系统,能不能共享同一套预测学习原理?\n\n## 核心思路:把潜状态拆成正交的预测因子\n\n论文的答案叫「正交预测分解」(Orthogonal Predictive Factorization,OPF)。它建立在联合嵌入预测架构(JEPA)之上,但把原本整块的潜空间目标拆解成一组互补的预测因子:每个因子走自己的专属预测通路去学,再在共享的预测设计里重组回完整状态,供下游读取、干预预测、规划和多步 rollout 使用。\n\nREADME 里的一句话概括得很清楚:不同领域保留各自的观测形式、上下文—目标构造和编码器,共享的是预测内核与潜状态接口——共性在原理层,不强求共用一套权重。这个设计直接决定了它「领域无关」的定位:换域只需要换观测适配器,预测核心不动。\n\n## 七个领域的实测账本\n\n评测覆盖视觉、生物、临床轨迹、控制、分子动力学、物理场、天气七个领域,实验类型包括表征学习、干预预测、分布外泛化和长时程动力学,具体包括 10 个匹配动力学任务、超过 1000 个临床事件的预测、以及四个系统各 100 步的分子 rollout。几个关键数字:\n\n- 对照匹配的 JEPA 基线,10 个动力学任务上的报告指标全部提升;\n- Interventional Pong 上,单次干预预测误差降低 34.8%;\n- 四个分子系统的单步与 100 步误差,在参比方法中均为最低;\n- 潜在轨道模态拟合出的开普勒缩放指数为 -1.4991。\n\n最有意思的是后两条。开普勒指数说明学到的潜在因子不只是拟合数字,还能对上物理定律;而由因子提名的一个生物干预,在细胞共培养、患者来源类器官、肿瘤碎片和小鼠实验中都获得了实验支持——这是「预测模型反哺实验科学」的完整闭环,也是这篇论文和普通刷榜文最大的差别。\n\n## 冷水:仓库里没有训练权重\n\n需要泼的冷水写在 GitHub 仓库的「Repository scope」一节:当前开源的是可复用的预测内核、任务设计工具和一个合成结构示例,**不包含论文里那些领域数据集和训练好的模型权重**,checkpoint manifest 记录的是一个未训练示例,明确声明不含性能主张。换句话说,想复现 34.8% 这类数字,数据集和训练流程都得自己搭。截至发稿,仓库 62 星、6 次提交、Apache-2.0 许可,Hugging Face 论文页 40 个 upvote——社区关注度刚起步。\n\n## 所以呢\n\n对做 AI for Science 的人,这篇的价值不在「又一个世界模型」,而在给「跨域共享预测原理」提供了一份可检验的工程蓝本:OPF 的因子化接口是模块化的,新域接入的成本被压到观测适配器一层。对更广泛的读者,它提示了一个方向——当下各家都在卷单域的生成质量,而「一套原理打通多个世界」的通用性主张,最终要靠湿实验和物理定律来仲裁,不是靠榜单。\n\n论文:https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.20800\n代码:https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FJEPA-Anything\n","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.20800","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},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"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},"a975e33b-7a69-413b-abc9-7db4843dbdfe","en","JEPA-Anything: One Predictive Core, Seven Worlds, Wet-Lab Backing","JEPA-Anything spans seven domains with one predictive core: all 10 dynamics tasks improve, intervention error falls 34.8%, and wet-lab tests agree.","The AI world talks a lot about world models, yet most predictive stacks remain siloed: one architecture for vision, another for molecular dynamics, yet another for clinical trajectories. JEPA-Anything (arXiv 2609.20800, published Sep 17) asks a more fundamental question — can cells, molecules, physical fields, and weather share a single predictive learning principle?\n\n## The core idea: factor the latent state orthogonally\n\nThe proposed answer is Orthogonal Predictive Factorization (OPF). Built on joint-embedding predictive architectures (JEPA), it replaces a monolithic latent target with a set of complementary predictive factors: each factor is learned through a dedicated pathway, then recombined within a shared predictive design into a complete state for downstream readout, intervention prediction, planning, and multi-step rollout.\n\nThe README puts it plainly: different domains keep their own observations, context–target construction, and encoders, while sharing the predictive core and the latent world-state interface. The commonality lives at the principle level — domains are not required to share a single encoder or one set of weights. That design choice is what makes the framework domain-agnostic: porting to a new domain means writing an observation adapter, not rebuilding the predictor.\n\n## The seven-domain scorecard\n\nEvaluation spans vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather, across representation learning, intervention prediction, out-of-distribution generalization, and long-horizon dynamics — including 10 matched dynamics tasks, forecasting of over 1,000 clinical events, and 100-step molecular rollouts across four systems. The key numbers:\n\n- Against matched JEPA baselines, reported metrics improve on all 10 dynamics tasks.\n- On Interventional Pong, single-intervention prediction error drops by 34.8%.\n- One-step and 100-step molecular errors are the lowest among compared methods in all four systems.\n- Latent orbital modes recover the Keplerian scaling exponent with a fitted slope of -1.4991.\n\nThe last two matter most. The Keplerian fit suggests the learned factors align with physical law rather than merely fitting numbers. And a factor-nominated biological intervention received experimental support in cell co-cultures, patient-derived organoids, tumor fragments, and mice — a full loop from predictive model back to bench experiment, which is what separates this paper from ordinary benchmark-chasing work.\n\n## Cold water: no trained weights in the repo\n\nThe GitHub repository's \"Repository scope\" section is explicit: the open-source release contains the reusable core, task-design tools, and a synthetic structural example — not the domain datasets or trained model weights from the paper. The checkpoint manifest records an untrained example with no performance claims. In other words, reproducing that 34.8% means assembling datasets and training pipelines yourself. As of writing, the repo sits at 62 stars, 6 commits, Apache-2.0, with 40 upvotes on the Hugging Face paper page — early-stage community traction.\n\n## So what\n\nFor AI-for-Science practitioners, the value is not \"another world model\" but a testable blueprint for cross-domain predictive learning: the factorized OPF interface is modular, and the cost of adding a domain is pushed down to the observation-adapter layer. For everyone else, it points at a direction — while the industry competes on per-domain generation quality, the universality claim of \"one principle, many worlds\" is ultimately arbitrated by wet experiments and physical laws, not leaderboards.\n\nPaper: https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.20800\nCode: https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FJEPA-Anything\n","jepa-anything-orthogonal-predictive-factorization","2026-09-19T23:10:37Z","2026-09-19T23:10:47.855205Z","2026-09-19T23:10:47.855215Z",true,"agent",114,[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},"d8bc7b5e-9eb0-475e-91b7-5a3390d2c6a6","2026年开源LLM爆发：Meta、阿里、Google竞相发布新一代模型","open-source-llm-boom-2026-q1-meta-alibaba-google","2026-04-24T04:06:08+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"c99751d5-418e-49d5-99d3-e43b84c80ec7","IBM与NASA开源月球基础模型:Lunar Foundation Model","nasa-ibm-lunar-foundation-model-sombench","2026-09-19T09:30:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"28c41f06-d20f-481c-b133-cd109af3aed1","答对之后停不下来:微软团队揪出在线蒸馏的 EOS 错配元凶","eos-mismatch-opd-length-inflation","2026-09-18T21:09:06+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"cb1e799d-d6d7-4ab9-9eaf-bea0aa432b06","Mistral 模型进驻 Firefox:119B 开放权重模型驱动浏览器 AI 助手","mistral-small-4-firefox-smart-window","2026-09-16T17:07:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"5535b4e4-21de-4ed6-a9bb-2b4d824e6568","F-Droid 一次更新的 102 款应用,72.5% 主要是 AI 写的","f-droid-72-percent-ai-written-audit","2026-09-15T15:12:16+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"dcd8b3e1-a3c7-4614-aba4-9002219ea5f6","LibreDB Studio 0.15 发布:本地 LLM 接管数据库交互","libredb-studio-local-llm-agent","2026-09-15T00:00:00+00:00"]