[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-small-tweak-to-robot-action-coding-boosts-learning-with-less-data":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},9681,"a-small-tweak-to-robot-action-coding-boosts-learning-with-less-data","A Small Tweak to Robot Action Coding Boosts Learning With Less Data","A new tokenization method lets robot AI models learn more from fewer demonstrations by randomizing how action sequences get encoded.","A new robot AI paper shows that how you encode actions into tokens matters almost as much as how many demonstrations you have.\n\nResearchers propose TOAST (TOkenization of Action sequences with STochastic sampling), a method for autoregressive vision-language-action models that predict robot actions as token sequences. The standard approach, called FAST, compresses continuous robot motions into a small set of discrete tokens, but it always picks one fixed tokenization for a given action. TOAST instead samples different, valid tokenizations of the same underlying motion during training, so the model sees varied token sequences that still decode to the same physical action. On the LIBERO benchmark, TOAST beat the deterministic FAST baseline by 6.8 points in success rate when trained on just 1\u002F16 of the available demonstrations, and it improved mean success rate by 15.8 points across four real robot manipulation tasks.\n\nCollecting robot demonstration data is slow and expensive, so any method that squeezes more learning out of fewer demos matters for teams building robot foundation models. The gains grow as data gets scarcer, which is exactly the situation facing labs and startups without warehouse-scale robot fleets.\n\nIt is a modest-sounding fix - randomize the labels, not the data - but that is the kind of unglamorous tweak that tends to get quietly folded into the next generation of robot policies.","[\"robotics\",\"ai\",\"machine-learning\",\"vision-language-action\"]","2026-10-02T04:00:00.000Z","2026-10-03T06:03:34.354Z","2026-10-03T06:03:38.657Z","published",null,[],"ai",[26,24,27,28],"robotics","machine-learning","vision-language-action",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.00899",0,{"sections":35},[36,39,43,47,52,56,60,65,70,75,80,85,90,95],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",5975,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",842,{"name":44,"slug":45,"count":46,"latest_published_at":18},"Policy","policy",438,{"name":48,"slug":49,"count":50,"latest_published_at":51},"Deals","deals",317,"2026-10-01T22:00:00.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":18},"Hardware","hardware",199,{"name":57,"slug":58,"count":59,"latest_published_at":18},"Science","science",173,{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",155,"2026-10-01T19:54:10.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Dev Tools","dev-tools",96,"2026-10-01T16:57:03.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",93,"2026-09-30T21:41:11.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Startups","startups",90,"2026-10-01T21:55:22.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",53,"2026-10-02T02:50:39.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"How-To","how-to",7,"2026-10-01T09:00:00.000Z"]