[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-researchers-optimize-how-ai-agents-pick-from-multiple-workflows":10,"sections":34},{"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":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},6511,"researchers-optimize-how-ai-agents-pick-from-multiple-workflows","Researchers Optimize How AI Agents Pick From Multiple Workflows","A new study shows that running several AI workflows at once and picking the best answer beats relying on a single top-performing workflow alone.","Running several AI reasoning workflows at once and picking the best result beats betting on a single workflow every time.\n\nThe researchers frame this as a \"portfolio-and-selector\" problem: a system runs several distinct workflows on the same task, then picks the final answer after seeing all the outputs. Extra runs can catch correct answers the single best workflow would have missed, but they cost more compute and add wrong-but-plausible answers that make the final selection harder. The team built exact formulations, linear-programming relaxations, and an ellipsoid-method search to figure out which mix of workflows, and how many runs, is worth the money. Tested on three benchmarks (ABCD, Schema-Guided Dialogue, and HotpotQA), the portfolio approach beat the best standalone workflow by 3.1, 7.5, and 0.9 percentage points respectively, with a workflow-generation method adding as much as 24.1 points on HotpotQA.\n\nThis matters because most agentic AI deployments still pick one workflow and hope it generalizes, treating extra inference runs as pure waste. This paper instead treats compute spend as a budget to allocate across workflow variety, closer to how a trading desk manages a portfolio than how most AI teams manage a single agent pipeline.\n\nThe catch is in the numbers: a 24.1-point jump on HotpotQA sits next to a 0.9-point gain on Schema-Guided Dialogue, with zero extra benefit from workflow generation on that same dataset. Running more workflows can help a lot, but how much depends entirely on the task, and someone still has to pay for all those extra runs.","[\"ai\",\"agentic-ai\",\"ai-research\",\"compute-costs\"]","2026-09-17T04:00:00.000Z","2026-09-17T21:38:05.490Z","2026-09-17T21:38:17.407Z","published",null,[],"ai",[24,26,27,28],"agentic-ai","ai-research","compute-costs",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.18126",0,{"sections":35},[36,40,44,49,54,58,62,67,72,76,81,86,91,96],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",3852,"2026-09-17T08:27:09.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",648,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",338,"2026-09-11T04:00:00.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":18},"Hardware","hardware",154,{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",114,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":18},"Dev Tools","dev-tools",73,{"name":77,"slug":78,"count":79,"latest_published_at":80},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",41,"2026-09-08T01:57:23.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]