{"id":7458,"date":"2025-09-22T04:29:00","date_gmt":"2025-09-21T19:29:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7458"},"modified":"2025-09-20T19:33:50","modified_gmt":"2025-09-20T10:33:50","slug":"7458","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7458","title":{"rendered":"Self-Improving Embodied Foundation Models\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Self-Improving Embodied Foundation Models\u00a0<\/strong>[21.8]<br>\u30ed\u30dc\u30c3\u30c8\u5de5\u5b66\u306b\u304a\u3051\u308b2\u6bb5\u968e\u306e\u30dd\u30b9\u30c8\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u624b\u6cd5\u3092\u63d0\u6848\u3059\u308b\u3002 \u7b2c1\u6bb5\u968e\u3067\u3042\u308bSupervised Fine-Tuning (SFT) \u306f\u3001a) \u884c\u52d5\u30af\u30ed\u30fc\u30cb\u30f3\u30b0\u3068b) \u30b9\u30c6\u30c3\u30d7\u30fb\u30c8\u30a5\u30fb\u30b4\u30fc\u306e\u4e88\u6e2c\u76ee\u7684\u306e\u4e21\u65b9\u3092\u7528\u3044\u305f\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30f3\u4e8b\u524d\u8a13\u7df4\u57fa\u790e\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u7b2c2\u6bb5\u968e\u3067\u306f\u3001\u30b9\u30c6\u30c3\u30d7\u30fb\u30c8\u30a5\u30fb\u30b4\u30fc\u4e88\u6e2c\u306b\u3088\u308a\u3001\u826f\u597d\u306a\u5f62\u72b6\u306e\u5831\u916c\u95a2\u6570\u3068\u5805\u7262\u306a\u6210\u529f\u691c\u51fa\u5668\u306e\u62bd\u51fa\u304c\u53ef\u80fd\u3068\u306a\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.15155v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.15155v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 18 Sep 2025 17:00:08 GMT)<\/li>\n\n\n\n<li>\u300c1) Supervised Fine-Tuning (SFT) wherein we fine-tune EFMs using behavioral cloning as well as \u201csteps-to-go\u201d prediction objectives, and 2) Self-Improvement (Online RL) wherein EFMs autonomously practice downstream tasks and rapidly improve via optimizing self-predicted rewards.\u300d\u3068\u3044\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u306e\u63d0\u6848\uff08EFM=\u00a0Embodied Foundation Models\uff09\u3002\u300cFinally, we demonstrated that this novel combination uniquely unlocks a capability not possible by current methods: autonomously aquiring new skills that generalize far beyond the tasks covered in the imitation learning datasets. These findings highlight the transformative potential of combining pretrained foundation models with online Self- Improvement to enable autonomous skill acquisition in robotics.\u300d\u3068\u52b9\u679c\u304c\u3042\u3063\u305f\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/sites.google.com\/view\/mfa-self-improvement\/home\">Anonymous Supplementary Videos for &#8220;On the Magic of Online Self-Improvement for Embodied Multimodal Foundation Models&#8221;\u00a0<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[356],"class_list":["post-7458","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-self-x"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7458","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=7458"}],"version-history":[{"count":2,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7458\/revisions"}],"predecessor-version":[{"id":7460,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7458\/revisions\/7460"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7458"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7458"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7458"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}