[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-ai-solver-keeps-physics-simulations-stable-over-long-runs":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},8092,"new-ai-solver-keeps-physics-simulations-stable-over-long-runs","New AI Solver Keeps Physics Simulations Stable Over Long Runs","A new neural network architecture cuts long-horizon simulation error nearly in half while using far less compute than existing physics solvers.","Researchers have built an AI system that keeps physics simulations accurate over long stretches of simulated time, a problem that has dogged neural alternatives to traditional solvers.\n\nThe team calls it the Latent Generative Solver, or LGS. It combines three pieces: a Physics VAE that compresses twelve different families of partial differential equations into one shared latent space, a transformer that generates the next step by flow matching instead of direct prediction, and a training trick called input noising that the authors back with a mathematical stability proof. Pretrained on 2.5 million trajectories across 16 physical systems, LGS matches the best deterministic baseline on single-step predictions, then pulls ahead on 15 of 16 systems by the 5- and 10-step mark. Over 20 steps, it cuts error from 56.1 percent down to 30.2 percent, while using 13 to 77 times less compute per step than recurrent rivals. It also adapted to a new, higher-resolution fluid-flow system it had never seen during pretraining, improving faster during finetuning than the U-AFNO baseline.\n\nThat combination is the interesting part. Deterministic solvers drift over long runs because small errors compound, while prior probabilistic solvers dodge that problem but usually only work for one type of equation. Handling twelve PDE families in one model, and generalizing to a new one with light finetuning, points toward simulators that do not need to be rebuilt for every new physical scenario.\n\nThe results are still confined to two-dimensional, regularly gridded systems and a fixed set of step counts, so the leap from arXiv preprint to your local climate model or engine simulator is not automatic.","[\"physics-simulation\",\"generative-ai\",\"machine-learning\",\"research\"]","2026-09-28T04:00:00.000Z","2026-09-28T09:07:15.233Z","2026-09-28T09:07:21.522Z","published",null,[],"ai",[26,27,28,29],"physics-simulation","generative-ai","machine-learning","research",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.11229",0,{"sections":36},[37,40,44,49,54,59,63,68,73,78,83,88,92,97],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",4791,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",762,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",399,"2026-09-27T18:39:02.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",261,"2026-09-27T15:30:35.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Hardware","hardware",188,"2026-09-27T20:46:36.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":18},"Science","science",151,{"name":64,"slug":65,"count":66,"latest_published_at":67},"Consumer Tech","consumer-tech",135,"2026-09-26T14:30:00.000Z",{"name":69,"slug":70,"count":71,"latest_published_at":72},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":74,"slug":75,"count":76,"latest_published_at":77},"Dev Tools","dev-tools",84,"2026-09-26T04:20:58.000Z",{"name":79,"slug":80,"count":81,"latest_published_at":82},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":84,"slug":85,"count":86,"latest_published_at":87},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":89,"slug":90,"count":86,"latest_published_at":91},"General","general","2026-09-26T17:02:42.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]