[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-give-ai-agents-different-evidence-and-forecasts-get-better":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},3395,"give-ai-agents-different-evidence-and-forecasts-get-better","Give AI Agents Different Evidence and Forecasts Get Better","A new framework shows that feeding every agent the same data causes groupthink — splitting the evidence pool cuts forecasting error by up to 18%.","When AI agents agree too fast, the crowd wisdom disappears.\n\nResearchers have released InfoDelphi, a multi-agent forecasting framework built around a deliberately simple idea: don't give every agent the same information. The system splits evidence into a shared public pool and disjoint private subsets, so each agent holds exclusive facts that can only surface through discussion with the others. The team tested it on PolyGym, a benchmark of 375 binary forecasting questions drawn from real-world prediction markets, and found InfoDelphi beat the best single-agent and multi-agent baselines by 12-18% on Brier score — a standard calibration metric — and by 4-8 percentage points in raw accuracy.\n\nThe finding matters because it names a specific failure mode that most multi-agent system designs have quietly ignored. When every agent reads the same briefing, deliberation becomes an echo chamber: the agents converge on the same priors before the first exchange, and the theoretical benefit of having multiple independent reasoners evaporates. The paper frames this as inter-agent error correlation — errors are no longer independent, so averaging doesn't help. Crucially, the researchers ran ablations stripping out the asymmetry and found it eliminated most of the accuracy gains, which pins the cause clearly.\n\nThis matters beyond forecasting. Multi-agent pipelines are spreading fast — coding assistants, research synthesizers, decision-support tools — and nearly all of them feed identical context to every agent. If the error-correlation argument generalizes, a lot of those systems are leaving meaningful quality on the table.\n\nThe uncomfortable footnote: the benchmark is 375 questions, and prediction markets have their own biases. Whether the gains hold outside carefully partitioned evidence settings is still an open question.","[\"ai\",\"multi-agent\",\"forecasting\",\"research\"]","2026-07-03T04:00:00.000Z","2026-07-03T04:58:06.053Z","2026-07-03T04:58:08.953Z","published",null,[],"ai",[24,26,27,28],"multi-agent","forecasting","research",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.01661",0,{"sections":35},[36,40,45,50,55,60,65,70,75,80,85,89,94,99],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":86,"slug":87,"count":83,"latest_published_at":88},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]