[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-exaone-demand-targets-retails-toughest-forecasting-problem":10,"sections":41},{"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":30,"tags":31,"sources":36,"feedback":40,"feedback_at":22,"cost_usd":40,"total_tokens":40},8190,"exaone-demand-targets-retails-toughest-forecasting-problem","EXAONE Demand Targets Retail's Toughest Forecasting Problem","A specialized forecasting model beats 36 general time-series AI systems on demand data, and real sales figures still edge out synthetic-only training.","A new AI model built just for demand forecasting claims to beat three dozen general-purpose rivals at their own game.\n\nThe model, called EXAONE Demand, is trained on a corpus of 11.3 million time series and 48.4 billion observations pulled from 73 sources, supplemented by a synthetic data generator that fills gaps in real-world demand records. Instead of retraining an entire foundation model, the researchers attached small low-rank adapter branches to a frozen general-purpose backbone, one branch each for four demand patterns: smooth, intermittent, erratic, and lumpy. A router reads eight statistics about an incoming series and decides how much weight each branch gets. The team built two versions, one trained on real and synthetic demand data together, the other on synthetic data alone, and tested both against 36 existing time-series foundation models across 22 held-out datasets.\n\nBoth EXAONE Demand versions beat all 36 rivals. That matters because demand data breaks most of the assumptions behind general-purpose time-series models: short histories, long stretches of zero sales, numbers censored by stock-outs, and swings driven by outside events, like promotions or weather, that the series itself never records. A model built around those quirks, rather than a bigger generic one, is more useful to retailers and supply chains than another all-purpose forecaster.\n\nThe version trained on real sales data plus synthetic data still beat the synthetic-only version, confirming that actual demand records add something a generator cannot fully replicate. That is a modest, honest result in a field that often oversells synthetic data as a free substitute for the real thing.","[\"demand-forecasting\",\"time-series-ai\",\"synthetic-data\",\"foundation-models\"]","2026-09-28T04:00:00.000Z","2026-09-28T16:12:55.564Z","2026-09-28T16:13:02.324Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Drop or hedge the closing claim that the synthetic-only version 'came close' to the real+synthetic version and call this 'the more interesting result' — the source only states real-world data 'adds a gain,' with no numbers to support how close the two versions actually were, so this reads as an invented quantitative characterization.","resolved","ai",[32,33,34,35],"demand-forecasting","time-series-ai","synthetic-data","foundation-models",[37],{"name":38,"url":39},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.30880",0,{"sections":42},[43,46,50,55,60,65,69,74,79,84,89,94,98,103],{"name":44,"slug":30,"count":45,"latest_published_at":18},"AI",4844,{"name":47,"slug":48,"count":49,"latest_published_at":18},"Security","security",762,{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",399,"2026-09-27T18:39:02.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Deals","deals",265,"2026-09-28T14:00:00.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Hardware","hardware",189,"2026-09-28T10:52:40.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":18},"Science","science",151,{"name":70,"slug":71,"count":72,"latest_published_at":73},"Consumer Tech","consumer-tech",135,"2026-09-26T14:30:00.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Dev Tools","dev-tools",84,"2026-09-26T04:20:58.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":95,"slug":96,"count":92,"latest_published_at":97},"General","general","2026-09-26T17:02:42.000Z",{"name":99,"slug":100,"count":101,"latest_published_at":102},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":104,"slug":105,"count":106,"latest_published_at":107},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]