
{"id":8700,"date":"2026-06-02T03:29:12","date_gmt":"2026-06-01T19:29:12","guid":{"rendered":"https:\/\/infernews.com\/blog\/scope-2\/"},"modified":"2026-06-02T03:29:12","modified_gmt":"2026-06-01T19:29:12","slug":"scope-2","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/scope-2\/","title":{"rendered":"\u7121\u9700\u5916\u90e8\u8cc7\u6599\u7684\u81ea\u535a\u5f08\u8a13\u7df4\uff1aSCOPE \u8b93\u6a21\u578b\u81ea\u5df1\u51fa\u984c\u81ea\u5df1\u7b54"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pasted-01ec96ac4b5e.jpg\" alt=\"Og image\"><\/figure>\n<p>\u73fe\u6709\u7684\u8a9e\u8a00\u6a21\u578b\u81ea\u535a\u5f08\uff08self-play\uff09\u8a13\u7df4\u65b9\u6cd5\uff0c\u5927\u591a\u53ea\u80fd\u8655\u7406\u6709\u660e\u78ba\u5c0d\u932f\u7684\u984c\u76ee\uff0c\u4f8b\u5982\u6578\u5b78\u904b\u7b97\u3002\u8607\u683c\u862d\u611b\u4e01\u5821\u5927\u5b78\u7b49\u6a5f\u69cb\u7684\u7814\u7a76\u5718\u968a\u63d0\u51fa\u4e86\u4e00\u500b\u540d\u70ba <strong>SCOPE<\/strong>\uff08Self-Play via Co-Evolving Policies\uff09\u7684\u6846\u67b6\uff0c\u628a\u81ea\u535a\u5f08\u62d3\u5c55\u5230\u6c92\u6709\u6a19\u6e96\u7b54\u6848\u7684\u958b\u653e\u5f0f\u4efb\u52d9\uff0c\u4f8b\u5982\u9700\u8981\u6574\u5408\u591a\u6bb5\u8cc7\u6599\u624d\u80fd\u5b8c\u6210\u7684\u554f\u7b54\u3002\u7814\u7a76\u5718\u968a\u4f86\u81ea University of Edinburgh\u3001Imperial College London \u8207 Miniml.AI\u3002<\/p>\n<p>SCOPE \u7684\u6838\u5fc3\u8a2d\u8a08\u662f\u8b93\u540c\u4e00\u500b\u57fa\u790e\u6a21\u578b\u5206\u8eab\u6210\u4e09\u500b\u89d2\u8272\uff1a<strong>Challenger<\/strong>\uff08\u51fa\u984c\u8005\uff09\u3001<strong>Solver<\/strong>\uff08\u7b54\u984c\u8005\uff09\u4ee5\u53ca <strong>Judge<\/strong>\uff08\u8a55\u5224\u8005\uff09\u3002Challenger \u8b80\u53d6\u4e00\u4efd\u6587\u4ef6\uff0c\u900f\u904e\u591a\u8f2a\u6aa2\u7d22\u5beb\u51fa\u96e3\u5ea6\u8cbc\u8fd1\u7b54\u984c\u8005\u6975\u9650\u7684\u984c\u76ee\uff1bSolver \u5247\u8981\u9760\u81ea\u5df1\u641c\u5c0b\u8cc7\u6599\u3001\u6574\u5408\u8b49\u64da\u5f8c\u4f5c\u7b54\uff1bJudge \u51cd\u7d50\u5728\u521d\u59cb\u72c0\u614b\uff0c\u6839\u64da\u540c\u4e00\u4efd\u6587\u4ef6\u64ec\u5b9a\u8a55\u5206\u6e96\u5247\uff0c\u4e26\u70ba\u6bcf\u9805\u6a19\u6e96\u7d66\u51fa\u56b4\u683c\u7684\u4e8c\u5143\u8a55\u5206\u3002\u4e09\u8005\u5b8c\u5168\u4e0d\u4f9d\u8cf4\u4eba\u5de5\u7de8\u5beb\u7684\u984c\u76ee\uff0c\u4e5f\u4e0d\u9700\u8981\u9ad4\u578b\u9f90\u5927\u7684\u524d\u6cbf\u6a21\u578b\u505a\u76e3\u7763\u3002<\/p>\n<p>\u9019\u500b\u6846\u67b6\u89e3\u6c7a\u4e86\u4e00\u500b\u95dc\u9375\u75db\u9ede\uff1a\u958b\u653e\u5f0f\u4efb\u52d9\u6c92\u6709\u56fa\u5b9a\u7b54\u6848\uff0c\u50b3\u7d71\u5f37\u5316\u5b78\u7fd2\u96e3\u4ee5\u7d66\u51fa\u53ef\u9760\u7684\u53cd\u994b\u3002SCOPE \u900f\u904e\u300c\u6587\u4ef6\u63a5\u5730\u300d\uff08document grounding\uff09\u88fd\u9020\u8cc7\u8a0a\u4e0d\u5c0d\u7a31\u2014\u2014Challenger \u548c Judge \u770b\u5f97\u5230\u539f\u6587\uff0cSolver \u770b\u4e0d\u5230\uff0c\u8feb\u4f7f\u7b54\u984c\u8005\u5fc5\u9808\u4e3b\u52d5\u6aa2\u7d22\u3002\u540c\u6642\uff0c\u984c\u76ee\u96e3\u5ea6\u88ab\u63a7\u5236\u5728\u7b54\u984c\u8005\u5f97\u5206\u7d04 50% \u7684\u4f4d\u7f6e\uff0c\u56e0\u70ba\u9019\u500b\u9ede\u7684\u53cd\u994b\u8b8a\u5316\u6700\u5927\uff0c\u6700\u6709\u5229\u65bc\u5b78\u7fd2\uff1b\u5f97\u5206\u4f4e\u65bc 0.2 \u6216\u9ad8\u65bc 0.8 \u7684\u984c\u76ee\u6703\u88ab\u904e\u6ffe\u6389\uff0c\u907f\u514d\u592a\u7c21\u55ae\u6216\u592a\u96e3\u7684\u5167\u5bb9\u6d6a\u8cbb\u8a13\u7df4\u8cc7\u6e90\u3002\u7814\u7a76\u4ea6\u52a0\u5165\u9577\u5ea6\u61f2\u7f70\u8207\u54c1\u8cea\u9580\u6abb\uff0c\u9632\u6b62\u6a21\u578b\u4ee5\u704c\u6c34\u6216\u6284\u539f\u6587\u7684\u65b9\u5f0f\u300c\u5237\u5206\u300d\u3002<\/p>\n<p>\u5728 Qwen2.5-7B \u7b49 7\u20138B \u898f\u6a21\u7684\u6a21\u578b\u4e0a\uff0cSCOPE \u5728 8 \u500b\u958b\u653e\u5f0f\u57fa\u6e96\u6e2c\u8a66\u4e2d\u6700\u9ad8\u53d6\u5f97 +10.4 \u5206\u7684\u63d0\u5347\uff0c\u6574\u9ad4\u589e\u5e45\u4ecb\u65bc +5.4 \u81f3 +10.4 \u5206\uff0c\u4e26\u5728 7 \u500b\u50b3\u7d71\u554f\u7b54\u57fa\u6e96\u4e0a\u540c\u6a23\u6709\u7a69\u5b9a\u9032\u6b65\uff0c\u904e\u7a0b\u4e2d\u4f7f\u7528\u4e86 0 \u689d\u4eba\u5de5\u7b56\u5283\u7684\u63d0\u793a\u3002\u5c0d\u65bc\u60f3\u4ee5\u6709\u9650\u9810\u7b97\u5fae\u8abf\u958b\u6e90\u6a21\u578b\u3001\u53c8\u8981\u517c\u9867\u958b\u653e\u5f0f\u751f\u6210\u54c1\u8cea\u7684\u958b\u767c\u8005\u8207\u7814\u7a76\u5718\u968a\uff0c\u9019\u500b\u65b9\u6cd5\u63d0\u4f9b\u4e86\u4e00\u689d\u4e0d\u4f9d\u8cf4\u5916\u90e8\u6578\u64da\u96c6\u7684\u8def\u5f91\u3002<\/p>\n<p><strong>\u9805\u76ee\uff1a<\/strong> <a href=\"https:\/\/edinburghnlp.github.io\/scope\/\" rel=\"noopener noreferrer\">https:\/\/edinburghnlp.github.io\/scope\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>SCOPE \u7528\u4e09\u500b\u89d2\u8272\u4e92\u76f8\u8a13\u7df4\uff0c\u628a\u300c\u6c92\u6709\u6a19\u6e96\u7b54\u6848\u300d\u7684\u958b\u653e\u5f0f\u4efb\u52d9\u4e5f\u5e36\u9032\u81ea\u535a\u5f08\u4e16\u754c\uff0c\u7121\u9700\u4eba\u5de5\u6578\u64da\u5373\u53ef\u63d0\u5347\u6a21\u578b\u8868\u73fe\u3002<\/p>\n","protected":false},"author":8,"featured_media":8699,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[133,127,197],"tags":[],"class_list":["post-8700","post","type-post","status-publish","format-standard","hentry","category-133","category-127","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8700","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=8700"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8700\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8699"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8700"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8700"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8700"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}