[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-vla-robots-learn-to-reuse-past-reasoning-not-redo-it":10,"sections":35},{"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":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},10911,"vla-robots-learn-to-reuse-past-reasoning-not-redo-it","VLA Robots Learn to Reuse Past Reasoning, Not Redo It","A new memory module called FLOWMEM lets robot control models recycle successful reasoning steps instead of recomputing them from scratch every time.","Robots that think through a task once shouldn't have to think it through again from zero the next time. A new system called FLOWMEM tries to fix that for vision-language-action models, the models that turn camera feeds and instructions into robot movements.\n\nHere's the actual mechanism. Most of these models do \"latent reasoning\": they build an internal representation of what's going on before deciding how to move, then throw that representation away once the action is taken. FLOWMEM instead saves fragments of past successful reasoning and retrieves the relevant ones as a new task unfolds, stitching them into a route that mirrors how the earlier success actually progressed. It then updates that borrowed route with fresh camera and sensor data before letting it drive the next action. In tests on two robot benchmarks, RoboMME and LIBERO-Plus, it hit 48.0% and 77.3% success, beating memory-free versions of the same approach by 1.7 and 4.1 percentage points.\n\n\n\nThose gains look modest on paper, but the idea matters more than the number. Most robot-control research has focused on making a single forward pass smarter. FLOWMEM is making the case that memory of what already worked is a cheaper and more reliable lever than better one-shot reasoning, which matters for anyone trying to run these models on real hardware with real latency budgets.\n\nA few percentage points of success rate won't win a trade show demo. Whether reused reasoning holds up outside curated benchmarks, where mistakes compound and environments drift, is the harder test nobody's run yet.","[\"robotics\",\"vision-language-action\",\"ai-research\",\"machine-learning\"]","2026-10-09T04:00:00.000Z","2026-10-09T21:10:09.743Z","2026-10-09T21:10:20.583Z","published",null,[],"ai",[26,27,28,29],"robotics","vision-language-action","ai-research","machine-learning",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.12090",0,{"sections":36},[37,40,44,49,54,58,62,67,72,77,82,87,92,97],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",6690,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",930,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":18},"Hardware","hardware",231,{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",192,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]