{"id":8135,"date":"2026-01-28T06:12:00","date_gmt":"2026-01-27T21:12:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8135"},"modified":"2026-01-24T21:17:40","modified_gmt":"2026-01-24T12:17:40","slug":"towards-execution-grounded-automated-ai-research","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8135","title":{"rendered":"Towards Execution-Grounded Automated AI Research"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Towards Execution-Grounded Automated AI Research\u00a0<\/strong>[106.9]<br>\u73fe\u5728\u306eLLM\u306f\u3057\u3070\u3057\u3070\u3082\u3063\u3068\u3082\u3089\u3057\u304f\u898b\u3048\u308b\u304c\u52b9\u679c\u306e\u306a\u3044\u30a2\u30a4\u30c7\u30a2\u3092\u751f\u6210\u3057\u307e\u3059\u3002\u5b9f\u884c\u57fa\u76e4\u5316\u306f\u5f79\u306b\u7acb\u3064\u304b\u3082\u3057\u308c\u306a\u3044\u304c\u3001\u81ea\u52d5\u5b9f\u884c\u304c\u5b9f\u73fe\u53ef\u80fd\u304b\u3069\u3046\u304b\u3001LLM\u304c\u5b9f\u884c\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u304b\u3089\u5b66\u3079\u308b\u304b\u3069\u3046\u304b\u306f\u4e0d\u660e\u3060\u3002 \u6211\u3005\u306f\u3001\u30a2\u30a4\u30c7\u30a2\u3092\u5b9f\u88c5\u3059\u308b\u81ea\u52d5\u5316\u30a8\u30b0\u30bc\u30ad\u30e5\u30fc\u30bf\u3092\u69cb\u7bc9\u3057\u3001\u305d\u306e\u6709\u52b9\u6027\u3092\u691c\u8a3c\u3059\u308b\u305f\u3081\u306b\u5927\u898f\u6a21\u306a\u4e26\u5217GPU\u5b9f\u9a13\u3092\u30ed\u30fc\u30f3\u30c1\u3059\u308b\u3002 \u672c\u7814\u7a76\u3067\u306f,\u9032\u5316\u7684\u63a2\u7d22\u3068\u5f37\u5316\u5b66\u7fd2\u3068\u3044\u3046,\u5b9f\u884c\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u304b\u3089\u5b66\u7fd2\u3059\u308b2\u3064\u306e\u65b9\u6cd5\u3092\u5206\u6790\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.14525v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.14525v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 20 Jan 2026 22:35:44 GMT)<\/li>\n\n\n\n<li>\u300cwe develop a large-scale automated idea executor system that can implement research ideas for open-ended and realistic research problems. Using this automated executor, we conduct an in-depth analysis of how well LLM ideators can learn from execution feedback to improve effectiveness through evolutionary search and RL. Execution- guided evolutionary search is sample-efficient and effective, but shows limited scaling. RL from execution reward suffers from diversity collapse and does not improve the upperbound.\u300d\u3068\u306e\u3053\u3068\u3067\u53ef\u80fd\u6027\u3068\u3068\u3082\u306b\u9650\u754c\u3092\u611f\u3058\u308b\u7d50\u679c<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/NoviScl\/Automated-AI-Researcher\">GitHub &#8211; NoviScl\/Automated-AI-Researcher<\/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":[623],"class_list":["post-8135","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-623"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8135","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=8135"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8135\/revisions"}],"predecessor-version":[{"id":8136,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8135\/revisions\/8136"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8135"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8135"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8135"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}