[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-model-pairs-a-fast-price-forecast-with-a-slower-news-check":10,"sections":40},{"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":35,"feedback":39,"feedback_at":22,"cost_usd":39,"total_tokens":39},8985,"new-model-pairs-a-fast-price-forecast-with-a-slower-news-check","New model pairs a fast price forecast with a slower news check","Researchers built a two-stage forecasting model that revises its quick numerical guess using news, beating rivals on longer-horizon tests.","A new AI forecasting system makes a quick numerical prediction, then pauses to reconsider itself using breaking news before locking in an answer.\n\nResearchers built DualCast, a forecasting framework that bolts a custom financial vocabulary onto a frozen Qwen3-8B language model, turning chunks of price history into learned tokens that capture both trend and finer shape details so patterns transfer across assets. A fast path trains only those new tokens to produce an initial forecast from the frozen model. A slower path then kicks in: a toggleable LoRA (low-rank adaptation) adapter takes that fast forecast plus news available at the time and revises the number, trained with reinforcement learning to reward real improvements over the initial guess. In tests on stocks and energy prices at intervals from five minutes to a week, the revised forecast was the most accurate in 8 of 12 scenarios, including every longest-horizon case.\n\nMost price-forecasting models either crunch numbers or read headlines, rarely both in the same pass. DualCast's two-step design is a workaround for a real tension: fast numerical models are cheap but blind to context, while text-heavy models are slow and prone to treating headlines as signal when they are noise. Letting a cheap model make the first guess and a pricier one only refine it is a sensible way to control costs, though the paper's own ablations show the news step helps inconsistently across markets.\n\nThat inconsistency is the real story here: before any trading desk wires this up, the open question is which markets the slow path actually helps, not whether language models can read a stock chart.","[\"ai\",\"llms\",\"finance\",\"forecasting\"]","2026-10-01T04:00:00.000Z","2026-10-01T14:23:20.744Z","2026-10-01T14:23:24.586Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"Define 'LoRA' on first use (e.g., 'a LoRA (low-rank adaptation) adapter') since the draft cites the acronym without ever expanding it.","resolved","ai",[30,32,33,34],"llms","finance","forecasting",[36],{"name":37,"url":38},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.38197",0,{"sections":41},[42,45,49,54,59,64,68,73,78,82,87,92,97,102],{"name":43,"slug":30,"count":44,"latest_published_at":18},"AI",5488,{"name":46,"slug":47,"count":48,"latest_published_at":18},"Security","security",809,{"name":50,"slug":51,"count":52,"latest_published_at":53},"Policy","policy",429,"2026-10-01T02:26:17.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":63},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":65,"slug":66,"count":67,"latest_published_at":18},"Science","science",162,{"name":69,"slug":70,"count":71,"latest_published_at":72},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":74,"slug":75,"count":76,"latest_published_at":77},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":79,"slug":80,"count":76,"latest_published_at":81},"Software","software","2026-09-30T21:41:11.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":103,"slug":104,"count":105,"latest_published_at":106},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]