[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-framework-fixes-optimization-models-from-user-overrides":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},9017,"new-framework-fixes-optimization-models-from-user-overrides","New Framework Fixes Optimization Models From User Overrides","A new technique called TACIT uses LLMs and optimization together to automatically patch flawed models using records of when humans overrode them.","A new system patches broken optimization models by learning from the decisions humans made instead of them.\n\nResearchers built TACIT, a framework that repairs misspecified optimization models - the kind used for scheduling, routing, and resource allocation - using historical records of past solutions and the overrides humans applied to them. It combines two methods: an LLM proposes new constraints or variables that optimization then calibrates and checks against the data, and separately, the optimization layer infers correction rules from the overrides, which the LLM turns into readable modeling constraints. The team tested the approach on 38 misspecification scenarios across nine classes of optimization problems, several pulled from real-world applications. TACIT fixed 78.9% of the broken models, compared with 60.5% for the best existing baseline.\n\nOptimization models are only as good as the assumptions baked into them, and what's usually missing is the stuff nobody wrote down - the tacit knowledge domain experts carry in their heads instead of in a spec sheet. Earlier fixes like inverse optimization and constraint learning tend to overfit on sparse data and produce formulations too tangled to trust. Splitting the labor - letting an LLM guess at plausible missing rules while optimization verifies them numerically - is a reasonable way to avoid both failure modes.\n\nA near-80% repair rate on 38 test cases is promising, but it is still a lab benchmark - the real test is whether this holds up against a planner who overrides the model for reasons even they cannot fully articulate.","[\"ai\",\"optimization\",\"llm-research\",\"arxiv\"]","2026-10-01T04:00:00.000Z","2026-10-01T15:53:10.771Z","2026-10-01T15:53:15.283Z","published",null,[],"ai",[24,26,27,28],"optimization","llm-research","arxiv",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.38434",0,{"sections":35},[36,39,43,48,53,58,62,67,72,76,81,86,91,96],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",5487,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",809,{"name":44,"slug":45,"count":46,"latest_published_at":47},"Policy","policy",429,"2026-10-01T02:26:17.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",162,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":73,"slug":74,"count":70,"latest_published_at":75},"Software","software","2026-09-30T21:41:11.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]