[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-agent-framework-helps-open-models-beat-gpt-6-on-benchmarks":10,"sections":35},{"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":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},7398,"new-agent-framework-helps-open-models-beat-gpt-6-on-benchmarks","New agent framework helps open models beat GPT-6 on benchmarks","A new framework lets agents learn an environment's quirks once, then reuse that memory to help open models beat GPT-6 on two benchmarks.","Researchers have built an AI agent framework that gets smarter without touching a model's weights.\n\nRSIAgent, detailed in a paper posted to arXiv, coordinates three roles - a curriculum agent, an actor agent, and a verifier agent - to explore a new digital environment on its own before ever attempting the task it was built for. It runs what the authors call a broad-then-deep strategy: first casting a wide net across parallel explorations to map the environment's structure, then digging into edge cases, hidden constraints, and the causal links between actions and their consequences. Everything it learns gets written into a memory store, which is then frozen and handed to the underlying model. There's no retraining or fine-tuning involved - just a reusable reference the model can consult.\n\nThat's the notable part: on the OSWorld-v2 and Agent's Last Exam benchmarks, this frozen memory alone was enough to help open-source models Kimi-K3 and GLM-5.3 outperform GPT-6, a frontier closed-source model. If that holds up, it suggests part of the gap between open and closed models isn't about raw capability - it's about how well an agent understands the quirks of the specific environment it's working in.\n\nTraining-free self-improvement has been pitched before as a cheap alternative to fine-tuning; the open question is whether a memory built by exploring one environment transfers to the next one, or has to be rebuilt from scratch every time.","[\"ai-agents\",\"open-source\",\"llm-benchmarks\",\"arxiv\"]","2026-09-23T04:00:00.000Z","2026-09-23T11:35:17.271Z","2026-09-23T11:35:22.865Z","published",null,[],"ai",[26,27,28,29],"ai-agents","open-source","llm-benchmarks","arxiv",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.15364",0,{"sections":36},[37,41,45,50,55,60,65,70,75,80,85,90,95,100],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",4347,"2026-09-23T12:00:00.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":18},"Security","security",713,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",371,"2026-09-23T12:00:43.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",211,"2026-09-23T13:00:46.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",170,"2026-09-23T11:59:23.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Science","science",134,"2026-09-23T09:00:00.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Consumer Tech","consumer-tech",110,"2026-09-22T20:00:00.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",81,"2026-09-23T09:56:13.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",79,"2026-09-22T22:21:13.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Startups","startups",65,"2026-09-22T22:06:48.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",45,"2026-09-22T15:35:06.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"General","general",43,"2026-09-21T23:48:56.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"Reviews","reviews",27,"2026-09-22T13:00:00.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]