[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-lean-becomes-a-teacher-not-just-a-grader":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},1677,"lean-becomes-a-teacher-not-just-a-grader","Lean Becomes a Teacher, Not Just a Grader","New research shows the Lean proof assistant can supply step-by-step feedback during AI training, not just a pass-fail verdict at the end.","AI models trained to write formal mathematical proofs are getting better feedback — mid-attempt, not just after.\n\nResearchers have found that Lean, a widely used formal proof assistant, can do more than check whether a completed proof is correct. During training, Lean can evaluate each individual proof step — called a tactic — and flag exactly where reasoning first goes wrong. The team parsed proof attempts into tactic sequences, used Lean's internal type-checking to mark which steps were locally sound, and fed that fine-grained signal back into a reinforcement learning objective. Models tested include STP-Lean and DeepSeek-Prover-V1.5, and the approach outperformed outcome-only training on benchmarks including MiniF2F and ProofNet.\n\nMost reinforcement learning from verifiable rewards works on a binary signal: right or wrong, full stop. Using a symbolic system as a step-level oracle is a meaningful shift, because it gives the model something closer to how a human tutor corrects work — pointing at the first bad line rather than just returning a failing grade. That density of feedback, grounded in formal logic rather than a neural model's judgment, is harder to game than softer reward signals.\n\nThe approach is narrow by design — formal proof verification is a domain where you can get reliable automated feedback at all, which is precisely why it works. The harder question is whether anything like this transfers to domains where no Lean-style oracle exists, and so far that answer remains open.","[\"ai\",\"formal-verification\",\"reinforcement-learning\",\"math\"]","2026-06-19T04:00:00.000Z","2026-06-19T09:45:00.319Z","2026-06-19T14:21:36.784Z","published",null,[],"ai",[24,26,27,28],"formal-verification","reinforcement-learning","math",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.20068",0,{"sections":35},[36,39,43,48,53,58,63,67,71,76,81,86,91,96],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",490,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",132,{"name":44,"slug":45,"count":46,"latest_published_at":47},"Policy","policy",88,"2026-06-16T09:26:09.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Consumer Tech","consumer-tech",78,"2026-06-16T17:58:24.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Hardware","hardware",62,"2026-06-18T15:24:16.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":62},"Deals","deals",58,"2026-06-19T14:43:50.000Z",{"name":64,"slug":65,"count":61,"latest_published_at":66},"Software","software","2026-06-16T20:00:00.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":18},"Dev Tools","dev-tools",50,{"name":72,"slug":73,"count":74,"latest_published_at":75},"Science","science",38,"2026-06-18T04:00:00.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Gaming","gaming",31,"2026-06-16T15:25:13.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"General","general",26,"2026-06-13T18:35:15.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"Startups","startups",23,"2026-06-16T15:00:00.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",19,"2026-06-14T08:00:00.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]