[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-system-lets-ai-evolve-its-own-machine-learning-code":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},6559,"new-system-lets-ai-evolve-its-own-machine-learning-code","New System Lets AI Evolve Its Own Machine Learning Code","A council of AI models drives evolutionary code generation that hit medal-level results on 19 of 22 MLE-bench Lite tasks, but leaned on human feedback.","A new research system evolves its own machine-learning code, grades the results, and tries again.\n\nResearchers describe Evolutionary Ensemble Search, or EES, as a pipeline where a \"council\" of role-specialized AI advisors turns task data and past results into search directions. An orchestrator hands those directions to execution specialists and an evolutionary engine, which mutates code, edits pipeline structure, and crosses successful variants to produce new candidate programs. Each new program has to actually run and generate its own validation evidence before it survives. The system keeps a memory of past runs and problem-specific lessons, and compatible candidates get combined in a validation-gated ensemble stage rather than just picking one winner.\n\nOn MLE-bench Lite, a public benchmark of 22 machine-learning tasks, the system hit medal-threshold performance on 19 of them, 86.36 percent of the set, with 11 golds, five silvers, and three bronzes across text, image, table, audio, and geometry problems. That is a strong score for automated pipeline-building. It also is not the full story: the campaign included human grading feedback between runs and routes to external sources, so this is a development result assembled with assistance, not proof of an unsupervised agent hitting those numbers alone.\n\nThat distinction matters more than the medal count. AutoML tools have promised to write and tune models for years; what is new here is treating the whole search process, memory, credit assignment, and mutation, as a persistent, versioned system rather than a one-off script. The paper's own framing, an \"achieved development result, not a blind autonomous-agent success rate,\" is more careful than most AI benchmark writeups manage.\n\nSomebody will strip out the human-in-the-loop parts and call this a fully autonomous data scientist. This paper does not.","[\"ai\",\"automl\",\"machine-learning\",\"benchmarks\"]","2026-09-17T04:00:00.000Z","2026-09-18T00:05:24.432Z","2026-09-18T00:05:36.358Z","published",null,[],"ai",[24,26,27,28],"automl","machine-learning","benchmarks",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.17590",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"]