[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-ai-framework-pushes-furniture-assembly-f1-to-628-percent":10,"sections":40},{"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":30,"tags":31,"sources":35,"feedback":39,"feedback_at":22,"cost_usd":39,"total_tokens":39},10822,"new-ai-framework-pushes-furniture-assembly-f1-to-628-percent","New AI Framework Pushes Furniture Assembly F1 to 62.8 Percent","A new zero-shot AI system tracks furniture manuals and physical states, lifting assembly-tree F1 from 38.58% to 62.80% over the prior best method.","A new AI framework can read furniture assembly manuals and track the physical state of a build well enough to beat the previous best system by a wide margin, though it still fumbles the physics.\n\nThe system, called AssemState, is a zero-shot framework that breaks diagram-heavy manuals into single-part operations using anchor-guided boundary assembly states, then reconstructs an assembly tree from them. It refines each part's 6D pose through iterative after-state feedback and checks the result with simulation-based release tests, essentially asking whether a part would actually stay put if physics were applied. Against the strongest prior baseline, F1 for assembly-tree recovery rose from 38.58% to 62.80%, and Tree Exact Match rose from 28.24% to 53.92%. On 243 separately evaluated part-level operations, the iterative refinement step pushed judge-accepted operations from 0% to 5.3% and cut mean Chamfer distance from 5.4111 to 1.7744.\n\nThe real story is in that last number. Furniture assembly is a useful stress test for whether multimodal language models can connect what they see on a page to what has to physically happen in 3D space, not just caption an image. A jump from 0% to 5.3% judge-accepted part placements shows these models can now parse structure, who attaches to whom, in what order, far better than before, but they are still mostly guessing when it comes to getting a single part's pose correct in the real world.\n\nWorth noting: this is virtual pose-checking, not a robot arm driving screws. The paper's own authors flag collisions, floating parts, mirror-orientation errors, and wrong-side attachments as ongoing failure modes. Call it progress on reading the instructions, not on building the bookshelf.","[\"ai\",\"robotics\",\"computer-vision\",\"research\"]","2026-10-08T04:00:00.000Z","2026-10-09T16:39:25.622Z","2026-10-09T16:39:31.947Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"The headline's 'Doubles Accuracy' claim doesn't match any single figure in the body (F1 rises 1.63x, Tree Exact Match rises 1.91x, part-acceptance goes from 0% to 5.3%) — tie the headline to one specific stated metric instead of a rounded-up 'doubles.'","resolved","ai",[30,32,33,34],"robotics","computer-vision","research",[36],{"name":37,"url":38},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.08446",0,{"sections":41},[42,46,50,55,60,65,69,74,79,84,89,94,99,104],{"name":43,"slug":30,"count":44,"latest_published_at":45},"AI",6609,"2026-10-09T04:00:00.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":45},"Security","security",926,{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Hardware","hardware",229,"2026-10-08T20:47:10.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":45},"Science","science",192,{"name":70,"slug":71,"count":72,"latest_published_at":73},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":105,"slug":106,"count":107,"latest_published_at":108},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]