[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-robot-ai-learns-to-use-unfamiliar-tools-by-watching-keypoints":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},4351,"robot-ai-learns-to-use-unfamiliar-tools-by-watching-keypoints","Robot AI Learns to Use Unfamiliar Tools by Watching Keypoints","A two-stage system called FORGE more than doubles robot success rates on novel tools by separating the reasoning from the doing.","Robots trained on one hammer do not know what to do with a rock — but a new approach from researchers aims to close that gap.\n\nA paper published this week introduces FORGE (FunctiOnal Reasoning and Grounded Execution), a two-stage robotic policy designed to handle tools it has never been trained on. The system first predicts how a tool should move — represented as 2D keypoint trajectories — using action-free data, then translates those predicted paths into actual robot motion using only a small number of demonstrations. Researchers tested FORGE on a seven-tool benchmark centered on hitting tasks, running trials in both simulation and the real world. It more than doubled the average success rate compared to leading existing methods.\n\nThe core insight is that visual similarity between tools — a rock and a hammer both look like things you swing — does not automatically translate into similar motor commands. FORGE sidesteps that mismatch by isolating functional reasoning into its own stage, one that does not need labeled robot action data to learn from. That decoupling is what lets it generalize: the system learns the intent of a tool category from video, then figures out execution separately.\n\nRobot manipulation research has long struggled with the sim-to-real gap and the data bottleneck — you need thousands of demonstrations per task, and every new tool restarts the clock. FORGE does not eliminate that problem, but a 2x success rate improvement on unseen tools, achieved with limited real-world demonstrations, is the kind of incremental result that quietly shifts what practitioners think is tractable. Whether it holds up outside hitting-function tasks — more varied object categories, less structured environments — is the next question worth watching.","[\"robotics\",\"ai\",\"manipulation\",\"computer-vision\"]","2026-07-08T04:00:00.000Z","2026-07-08T06:53:03.885Z","2026-07-08T06:53:07.546Z","published",null,[],"ai",[26,24,27,28],"robotics","manipulation","computer-vision",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.05780",0,{"sections":35},[36,40,45,50,55,60,65,70,75,80,85,89,94,99],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":86,"slug":87,"count":83,"latest_published_at":88},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]