[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-ai-ad-agents-learn-faster-when-rl-is-targeted-not-blanket":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},7440,"ai-ad-agents-learn-faster-when-rl-is-targeted-not-blanket","AI Ad Agents Learn Faster When RL Is Targeted, Not Blanket","A study on AI advertising agents finds targeted reinforcement learning cuts data leakage while using less compute than applying RL everywhere.","A new study says the smartest way to train AI advertising agents with reinforcement learning is to barely use it.\n\nResearchers tested \"enterprise analytics agents\": AI systems that pull business data, call APIs, and run code across many steps to handle advertising tasks. They compared supervised fine-tuning (SFT), where the model imitates example transcripts, against reinforcement learning (RL), where it gets rewarded for good outcomes. Rather than applying RL everywhere, they sorted each skill into one of three buckets (skills SFT already nailed, skills both stages improved, and skills only RL could unlock) and routed training accordingly. That diagnostic correctly predicted the outcome in 15 of 18 follow-up experiments, and on a GPT-OSS 120B model the targeted approach beat a baseline \"frontier Control\" on seven of eight advertiser skills, with the biggest gain on a non-disclosure skill at +11.27 points.\n\nFor agents handling advertiser data, the real payoff showed up in what they stopped leaking. An outside audit found the targeted RL approach cut standard data leakage from 11.8% to 2.9% and adversarial leakage from 22.9% to 6.8%, without sacrificing usefulness, since actionability barely moved, from 86.2% to 85.7%. Against applying RL uniformly across all skills, targeting it more than doubled the average performance gain, from +1.62 to +3.57 points, while using 43% less RL compute.\n\nIt's a reminder that for agents built to handle other people's data, teaching them what not to say can matter as much as teaching them what to do.","[\"ai\",\"reinforcement-learning\",\"ai-agents\",\"advertising\"]","2026-09-23T04:00:00.000Z","2026-09-23T13:30:58.938Z","2026-09-23T13:31:03.096Z","published",null,[],"ai",[24,26,27,28],"reinforcement-learning","ai-agents","advertising",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.22194",0,{"sections":35},[36,40,44,49,54,59,64,69,74,79,84,89,94,99],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",4347,"2026-09-23T12:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",713,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",371,"2026-09-23T12:00:43.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",212,"2026-09-23T13:00:46.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Hardware","hardware",170,"2026-09-23T11:59:23.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":63},"Science","science",134,"2026-09-23T09:00:00.000Z",{"name":65,"slug":66,"count":67,"latest_published_at":68},"Consumer Tech","consumer-tech",110,"2026-09-22T20:00:00.000Z",{"name":70,"slug":71,"count":72,"latest_published_at":73},"Software","software",81,"2026-09-23T09:56:13.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Dev Tools","dev-tools",79,"2026-09-22T22:21:13.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",65,"2026-09-22T22:06:48.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",45,"2026-09-22T15:35:06.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",43,"2026-09-21T23:48:56.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",27,"2026-09-22T13:00:00.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]