{"id":7888,"date":"2025-12-11T05:40:00","date_gmt":"2025-12-10T20:40:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7888"},"modified":"2025-12-06T17:43:52","modified_gmt":"2025-12-06T08:43:52","slug":"guided-self-evolving-llms-with-minimal-human-supervision","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7888","title":{"rendered":"Guided Self-Evolving LLMs with Minimal Human Supervision"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Guided Self-Evolving LLMs with Minimal Human Supervision\u00a0<\/strong>[53.1]<br>\u7121\u8a98\u5c0e\u306e\u81ea\u5df1\u9032\u5316\u30b7\u30b9\u30c6\u30e0\u306f\u3001\u3057\u3070\u3057\u3070\u8a13\u7df4\u3068\u3057\u3066\u7d20\u65e9\u304f\u3001\u307e\u305f\u306f\u52a3\u5316\u3059\u308b\u3002 R-Few\u306f\u30ac\u30a4\u30c9\u4ed8\u304d\u30bb\u30eb\u30d5\u30d7\u30ec\u30a4\u30c1\u30e3\u30ec\u30f3\u30b8\u30e3\u30fc(Self-Play Challenger)\u8cb7\u53ce\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3001\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u5185\u63a5\u5730\u3068\u6df7\u5408\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3092\u901a\u3058\u3066\u3001\u8efd\u91cf\u306a\u4eba\u9593\u306e\u76e3\u8996\u3092\u53d6\u308a\u5165\u308c\u3066\u3044\u308b\u3002 R-Few\u306f\u3001\u6570\u5b66\u3068\u4e00\u822c\u7684\u306a\u63a8\u8ad6\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u4e00\u8cab\u3057\u305f\u53cd\u5fa9\u7684\u306a\u6539\u5584\u3092\u5b9f\u73fe\u3057\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2512.02472v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2512.02472v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 02 Dec 2025 07:06:11 GMT)<\/li>\n\n\n\n<li>\u300cwe introduce R-FEW, a guided Self-Play Challenger\u2013Solver framework that incorporates lightweight human oversight through in-context grounding and mixed training. At each iteration, the Challenger samples a small set of human-labeled examples to guide synthetic ques- tion generation, while the Solver jointly trains on human and synthetic examples under an online, difficulty-based curriculum. Across math and general reasoning benchmarks, R-Few achieves consistent and iterative improvements.\u300d\u3068\u6700\u8fd1\u3088\u304f\u898b\u308b\u5171\u540c\u9032\u5316\u7684\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u63d0\u6848\u3002\u3068\u3066\u3082\u6d41\u884c\u3063\u3066\u3044\u3066\u6709\u52b9\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u306a\u306e\u3060\u3068\u601d\u3046\u3002<\/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-7888","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\/7888","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=7888"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7888\/revisions"}],"predecessor-version":[{"id":7889,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7888\/revisions\/7889"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7888"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7888"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7888"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}