[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-ai-agents-still-struggle-to-rebuild-slides-from-diagrams":10,"sections":33},{"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":28,"feedback":32,"feedback_at":22,"cost_usd":32,"total_tokens":32},7898,"ai-agents-still-struggle-to-rebuild-slides-from-diagrams","AI Agents Still Struggle to Rebuild Slides From Diagrams","A new benchmark called PPTBench found top AI agents still fail to accurately rebuild scientific diagrams as editable slides.","AI agents can already build webpages and games from scratch, but redrawing a PowerPoint slide still trips them up.\n\nA new benchmark called PPTBench tests whether AI coding agents can look at a diagram and rebuild it as an editable slide, not just a static image, using 500 scientific flow diagrams pulled from real arXiv papers. Each agent must reconstruct its diagram as a single PPTX page made of native, editable objects: shapes, text boxes, and arrows, which a four-stage \"Agentic Judge\" then checks for file validity, semantic accuracy, rendering quality, and fine visual detail. Across 31 setups spanning different model families, reasoning effort levels, and agent harnesses, the best configuration, built on Kimi K3, scored just 67.80 out of 100, and the median score across all setups was 19.47.\n\nThe gap between a file that opens and a file that's actually right is the real story: agents reliably produced valid PPTX files but consistently botched semantic and visual details, especially text. More reasoning mainly helped agents clear basic pass or fail gates rather than nail the details, while stronger verification during generation tracked more consistently with higher quality results.\n\nSo if you're picturing an AI that turns your rough sketch into a polished deck unsupervised, don't get too excited. It can produce a file that opens in PowerPoint. Whether it looks like what you actually drew is still closer to a coin flip than a sure thing.","[\"ai\",\"benchmarks\",\"coding-agents\"]","2026-09-25T04:00:00.000Z","2026-09-26T05:32:09.220Z","2026-09-26T05:32:14.980Z","published",null,[],"ai",[24,26,27],"benchmarks","coding-agents",[29],{"name":30,"url":31},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.29718",0,{"sections":34},[35,39,44,49,54,59,64,69,74,79,84,89,94,99],{"name":36,"slug":24,"count":37,"latest_published_at":38},"AI",4612,"2026-09-25T21:57:05.000Z",{"name":40,"slug":41,"count":42,"latest_published_at":43},"Security","security",744,"2026-09-25T21:09:27.000Z",{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",392,"2026-09-25T18:44:30.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",256,"2026-09-25T17:00:53.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Hardware","hardware",185,"2026-09-25T15:00:22.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":63},"Science","science",142,"2026-09-25T14:07:46.000Z",{"name":65,"slug":66,"count":67,"latest_published_at":68},"Consumer Tech","consumer-tech",132,"2026-09-25T15:30:00.000Z",{"name":70,"slug":71,"count":72,"latest_published_at":73},"Software","software",90,"2026-09-25T20:55:00.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Dev Tools","dev-tools",82,"2026-09-25T09:59:40.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",76,"2026-09-25T18:33:59.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",46,"2026-09-25T02:12:57.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",30,"2026-09-24T20:07:31.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]