{"id":30351,"date":"2026-10-12T10:28:03","date_gmt":"2026-10-12T02:28:03","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=30351"},"modified":"2026-10-14T18:39:57","modified_gmt":"2026-10-14T10:39:57","slug":"%e5%91%8a%e5%88%ab-agent-%e8%87%aa%e8%bf%9b%e5%8c%96%e8%bf%87%e6%8b%9f%e5%90%88%ef%bc%81%e8%b0%b7%e6%ad%8c-rrsi%ef%bc%9a%e8%ae%a9-ai-%e6%99%ba%e8%83%bd%e4%bd%93%e5%ae%9e%e7%8e%b0%e7%a8%b3%e5%ae%9a","status":"publish","type":"post","link":"https:\/\/www.ixpower.cn\/index.php\/2026\/10\/12\/30351\/","title":{"rendered":"\u544a\u522b Agent \u81ea\u8fdb\u5316\u8fc7\u62df\u5408\uff01\u8c37\u6b4c RRSI\uff1a\u8ba9 AI \u667a\u80fd\u4f53\u5b9e\u73b0\u7a33\u5b9a\u9012\u5f52\u81ea\u6211\u6539\u8fdb"},"content":{"rendered":"<p>RRSI \u7684\u5168\u79f0\u662f\u00a0Regularized Recursive Self-Improvement of Agent Harnesses\uff0c\u5373\u667a\u80fd\u4f53\u6846\u67b6\u6b63\u5219\u5316\u9012\u5f52\u81ea\u6211\u6539\u8fdb\u3002<\/p>\n<p>\u5b83\u662f\u8c37\u6b4c\u7814\u7a76\u9662\u63a8\u51fa\u7684\u65b0\u4e00\u4ee3\u00a0AI\u00a0\u667a\u80fd\u4f53\u81ea\u8fdb\u5316\u6280\u672f\uff0c\u53ef\u4ee5\u89e3\u51b3\u4f20\u7edf\u00a0Agent\u00a0\u81ea\u6211\u8fed\u4ee3\u8fc7\u7a0b\u4e2d\u666e\u904d\u5b58\u5728\u7684\u8fc7\u62df\u5408\u3001\u5237\u5206\u3001\u6cdb\u5316\u80fd\u529b\u9000\u5316\u3001\u7b97\u529b\u5197\u4f59\u7b49\u884c\u4e1a\u75db\u70b9\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s2.51cto.com\/oss\/202610\/12\/36b020c631081f5421f2384a72061f3a36c935.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u533a\u522b\u4e8e\u9700\u8981\u4fee\u6539\u5927\u6a21\u578b\u6743\u91cd\u7684\u9ad8\u9636\u81ea\u8fdb\u5316\u65b9\u6848\uff0c<strong>RRSI\u00a0\u5168\u7a0b<\/strong><strong>\u51bb\u7ed3\u57fa\u7840\u5927\u6a21\u578b<\/strong><strong>\uff0c\u4ec5\u9488\u5bf9\u667a\u80fd\u4f53\u5916\u5c42\u8fd0\u884c\u6846\u67b6\uff08Harness\uff09\u8fdb\u884c\u8fed\u4ee3\u4f18\u5316<\/strong>\uff0c\u5305\u62ec\u63d0\u793a\u8bcd\u3001\u5de5\u5177\u8c03\u7528\u89c4\u5219\u3001\u8bb0\u5fc6\u673a\u5236\u3001\u4e0a\u4e0b\u6587\u7ba1\u7406\u3001\u63a7\u5236\u6d41\u3001\u5b50\u667a\u80fd\u4f53\u914d\u7f6e\u7b49\u6838\u5fc3\u6a21\u5757\u3002\u8fd9\u79cd\u8f7b\u91cf\u5316\u8fed\u4ee3\u65b9\u5f0f\u95e8\u69db\u66f4\u4f4e\u3001\u5b89\u5168\u6027\u66f4\u9ad8\uff0c\u9002\u914d\u7edd\u5927\u591a\u6570\u843d\u5730\u573a\u666f\u3002<\/p>\n<p>RRSI\u00a0\u6700\u6838\u5fc3\u7684\u521b\u65b0\u662f<strong>\u5f15\u5165\u6b63\u5219\u5316\u7ea6\u675f\u673a\u5236<\/strong>\u3002\u4f20\u7edf\u667a\u80fd\u4f53\u81ea\u6211\u6539\u8fdb\u5c5e\u4e8e\u00a0\u201c\u65e0\u7ea6\u675f\u8d2a\u5fc3\u8fed\u4ee3\u201d\uff0c\u53ea\u8981\u5728\u6d4b\u8bd5\u4efb\u52a1\u4e0a\u5206\u6570\u63d0\u5347\u5c31\u4fdd\u7559\u4fee\u6539\uff0c\u6781\u6613\u5bfc\u81f4\u6a21\u578b\u9002\u914d\u6d4b\u8bd5\u96c6\u4e13\u5c5e\u89c4\u5f8b\u3001\u8bb0\u5fc6\u7b54\u6848\uff0c\u6700\u7ec8\u51fa\u73b0\u00a0\u201c\u8bad\u7ec3\u5206\u6570\u66b4\u6da8\u3001\u771f\u5b9e\u573a\u666f\u62c9\u80ef\u201d\u00a0\u7684\u8fc7\u62df\u5408\u95ee\u9898\u3002<\/p>\n<p>\u800c\u00a0<strong>RRSI\u00a0\u4f1a\u5bf9\u6bcf\u4e00\u6b21\u667a\u80fd\u4f53\u81ea\u6211\u4fee\u6539\u7684\u65b9\u6848\u8fdb\u884c\u6821\u9a8c\u3001\u7b5b\u9009\u3001\u526a\u679d\uff0c\u675c\u7edd\u65e0\u6548\u4f18\u5316\u3001\u865a\u5047\u63d0\u5347\u548c\u7b97\u529b\u6d6a\u8d39\u3002<\/strong><\/p>\n<p>\u4ece\u6280\u672f\u843d\u5730\u5f62\u6001\u6765\u770b\uff0cRRSI\u00a0\u5b8c\u6574\u8fd0\u884c\u94fe\u8def\u4f9d\u6258\u00a0Vertex AI\u00a0\u5b8c\u6210\u8349\u7a3f\u6da6\u8272\u3001Claude Opus\u00a0\u8f85\u52a9\u4f18\u5316\u3001Docker\u00a0\u57fa\u51c6\u8bc4\u5206\uff0c\u6838\u5fc3\u7684\u51b3\u7b56\u7b5b\u9009\u903b\u8f91\u5219\u901a\u8fc7\u7eaf\u00a0Python\u00a0\u4ee3\u7801\u5b9e\u73b0\uff0c\u5f00\u53d1\u8005\u53ef\u76f4\u63a5\u672c\u5730\u5316\u8c03\u7528\u3001\u4e8c\u6b21\u5f00\u53d1\uff0c\u65e0\u9700\u590d\u6742\u7684\u6a21\u578b\u5fae\u8c03\u6210\u672c\u3002<\/p>\n<p>\u5176\u6838\u5fc3\u6280\u672f\u6a21\u5757\u5305\u542b\uff1a\u5206\u6570\u4f30\u8ba1\u5668\u3001\u6821\u51c6\u566a\u58f0\u5e26\u3001\u53cc\u5206\u652f\u9009\u62e9\u7b97\u6cd5\u3001\u9000\u706b\u4fee\u6539\u9884\u7b97\u3001\u786e\u5b9a\u6027\u6cc4\u6f0f\u68c0\u6d4b\u3001\u8fed\u4ee3\u5386\u53f2\u8bb0\u5f55\u7ba1\u7406\u516d\u5927\u6838\u5fc3\u7ec4\u4ef6\uff0c\u5171\u540c\u6784\u6210\u95ed\u73af\u3001\u53ef\u63a7\u3001\u53ef\u8ffd\u6eaf\u7684\u667a\u80fd\u4f53\u81ea\u8fdb\u5316\u4f53\u7cfb\u3002<\/p>\n<h2>RRSI\u00a0\u4e0e\u00a0AI Agent\u00a0\u7684\u6838\u5fc3\u5173\u8054<\/h2>\n<p>\u60f3\u8981\u8bfb\u61c2\u00a0RRSI\uff0c\u9996\u5148\u8981\u7406\u6e05\u00a0Agent\u00a0\u548cHarness\u00a0\u7684\u6838\u5fc3\u5173\u7cfb\uff0c\u8fd9\u4e5f\u662f\u00a0RRSI\u00a0\u6280\u672f\u843d\u5730\u7684\u524d\u63d0\u3002<\/p>\n<p>\u7b80\u5355\u6765\u8bf4\uff1a<strong>Agent\u00a0\u662f\u6267\u884c\u4efb\u52a1\u7684\u4e3b\u4f53\uff0cHarness\u00a0\u662f\u7ea6\u675f\u548c\u652f\u6491\u4e3b\u4f53\u8fd0\u884c\u7684\u89c4\u5219\u9aa8\u67b6<\/strong>\u3002<\/p>\n<p>\u5927\u6a21\u578b\u662f\u667a\u80fd\u4f53\u7684\u00a0\u201c\u5927\u8111\u201d\uff0cHarness\u00a0\u662f\u667a\u80fd\u4f53\u7684\u00a0\u201c\u795e\u7ecf\u4e0e\u9aa8\u9abc\u201d\uff0c\u51b3\u5b9a\u5927\u8111\u7684\u6307\u4ee4\u5982\u4f55\u843d\u5730\u3001\u5de5\u5177\u5982\u4f55\u8c03\u7528\u3001\u8bb0\u5fc6\u5982\u4f55\u5b58\u50a8\u3001\u8fed\u4ee3\u5982\u4f55\u8fdb\u884c\u3002\u800c\u00a0RRSI\u00a0\u5c31\u662f\u4e13\u95e8\u4f18\u5316\u8fd9\u5957\u00a0\u201c\u9aa8\u9abc\u7cfb\u7edf\u201d\u00a0\u7684\u8fed\u4ee3\u7b97\u6cd5\u3002<\/p>\n<p>\u4e24\u8005\u7684\u6df1\u5ea6\u7ed1\u5b9a\u5173\u7cfb\u4e3b\u8981\u4f53\u73b0\u5728\u4e09\u70b9\uff1a<\/p>\n<h4>1.\u00a0\u8fed\u4ee3\u8f7d\u4f53\u7ed1\u5b9a\uff1aRRSI\u00a0\u4ec5\u4f18\u5316\u00a0Agent\u00a0\u5916\u5c42\u6846\u67b6<\/h4>\n<p>\u4f20\u7edf\u00a0RSI\uff08\u9012\u5f52\u81ea\u6211\u6539\u8fdb\uff09\u8bd5\u56fe\u4fee\u6539\u6a21\u578b\u6743\u91cd\uff0c\u8fed\u4ee3\u6210\u672c\u9ad8\u3001\u98ce\u9669\u5927\u3001\u6781\u6613\u5d29\u574f\u3002<\/p>\n<p>\u800c\u00a0RRSI\u00a0\u5b8c\u5168\u805a\u7126\u00a0Agent\u00a0\u7684\u00a0Harness\u00a0\u5c42\uff0c\u4e0d\u89e6\u78b0\u57fa\u7840\u6a21\u578b\u53c2\u6570\uff0c\u901a\u8fc7\u4f18\u5316\u63d0\u793a\u8bcd\u3001\u4e0a\u4e0b\u6587\u538b\u7f29\u3001\u5de5\u5177\u8c03\u5ea6\u3001\u5931\u8d25\u91cd\u8bd5\u673a\u5236\u7b49\u5de5\u7a0b\u5c42\u903b\u8f91\uff0c\u5b9e\u73b0\u667a\u80fd\u4f53\u80fd\u529b\u5347\u7ea7\uff0c\u9002\u914d\u6240\u6709\u4e3b\u6d41\u5927\u6a21\u578b\u667a\u80fd\u4f53\u3002<\/p>\n<h4>2.\u00a0\u8fed\u4ee3\u903b\u8f91\u8d4b\u80fd\uff1a\u89c4\u8303\u00a0Agent\u00a0\u81ea\u6211\u8fdb\u5316\u5168\u6d41\u7a0b<\/h4>\n<p>\u666e\u901a\u00a0Agent\u00a0\u7684\u81ea\u6211\u8fed\u4ee3\u662f\u5355\u5411\u8d2a\u5fc3\u903b\u8f91\uff1a\u6267\u884c\u4efb\u52a1\u2192\u590d\u76d8\u9519\u8bef\u2192\u4fee\u6539\u914d\u7f6e\u2192\u4fdd\u7559\u9ad8\u5206\u65b9\u6848\u3002<\/p>\n<p>RRSI\u00a0\u4e3a\u8fd9\u5957\u6d41\u7a0b\u589e\u52a0\u4e86\u6b63\u5219\u5316\u7ea6\u675f\uff0c\u4ece\u00a0\u201c\u53ea\u8981\u6da8\u5206\u5c31\u4fdd\u7559\u201d\u00a0\u5347\u7ea7\u4e3a\u00a0\u201c\u5fc5\u987b\u8bc1\u660e\u6cdb\u5316\u4ef7\u503c\u3001\u7b97\u529b\u6027\u4ef7\u6bd4\u3001\u65e0\u6f0f\u6d1e\u98ce\u9669\u624d\u80fd\u4fdd\u7559\u201d\uff0c\u8ba9\u00a0Agent\u00a0\u7684\u6bcf\u4e00\u6b21\u8fdb\u5316\u90fd\u771f\u5b9e\u6709\u6548\u3002<\/p>\n<h4>3.\u00a0\u80fd\u529b\u4e0a\u9650\u7a81\u7834\uff1a\u89e3\u51b3\u00a0Agent\u00a0\u8fed\u4ee3\u56fa\u6709\u7f3a\u9677<\/h4>\n<p>\u65e0\u7ea6\u675f\u8fed\u4ee3\u7684\u00a0Agent\u00a0\u4f1a\u9677\u5165\u4e09\u5927\u56f0\u5883\uff1a\u62df\u5408\u6d4b\u8bd5\u96c6\u89c4\u5f8b\u3001\u8bef\u5c06\u8bc4\u4f30\u566a\u58f0\u5f53\u4f5c\u80fd\u529b\u63d0\u5347\u3001\u4e0d\u65ad\u5806\u53e0\u5197\u4f59\u903b\u8f91\u5bfc\u81f4\u7b97\u529b\u66b4\u6da8\u3002<\/p>\n<p>RRSI\u00a0\u901a\u8fc7\u566a\u58f0\u6821\u51c6\u3001\u6210\u672c\u7ea6\u675f\u3001\u7ed3\u6784\u526a\u679d\u3001\u6cc4\u6f0f\u68c0\u6d4b\u56db\u5927\u673a\u5236\uff0c\u8ba9\u00a0Agent\u00a0\u8fed\u4ee3\u4ece\u00a0\u201c\u7247\u9762\u5237\u5206\u201d\u00a0\u8f6c\u5411\u00a0\u201c\u901a\u7528\u80fd\u529b\u63d0\u5347\u201d\uff0c\u5927\u5e45\u63d0\u5347\u8de8\u573a\u666f\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n<h2>\u4f20\u7edf\u00a0Agent\u00a0\u81ea\u8fdb\u5316\u9762\u4e34\u7684\u6280\u672f\u6311\u6218<\/h2>\n<p>\u5728\u00a0RRSI\u00a0\u95ee\u4e16\u524d\uff0c\u884c\u4e1a\u5185\u4e3b\u6d41\u7684\u667a\u80fd\u4f53\u9012\u5f52\u81ea\u6211\u6539\u8fdb\u65b9\u6848\uff0c\u666e\u904d\u5b58\u5728\u56db\u5927\u65e0\u6cd5\u89c4\u907f\u7684\u6280\u672f\u77ed\u677f\uff0c\u4e5f\u662f\u5236\u7ea6\u00a0Agent\u00a0\u843d\u5730\u89c4\u6a21\u5316\u8fed\u4ee3\u7684\u4e3b\u8981\u74f6\u9888\uff1a<\/p>\n<h4>1.\u00a0\u4e25\u91cd\u7684\u6d4b\u8bd5\u96c6\u8fc7\u62df\u5408<\/h4>\n<p>\u667a\u80fd\u4f53\u53cd\u590d\u5728\u540c\u4e00\u6279\u6f14\u5316\u4efb\u52a1\uff08Evolve Set\uff09\u4e0a\u8fed\u4ee3\u4f18\u5316\uff0c\u4f1a\u9010\u6e10\u8bb0\u5fc6\u4efb\u52a1\u4e13\u5c5e\u89e3\u9898\u5957\u8def\u3001\u56fa\u5b9a\u7b54\u6848\u3001\u6570\u636e\u7279\u5f81\uff0c\u800c\u975e\u5b66\u4e60\u901a\u7528\u6267\u884c\u903b\u8f91\u3002<\/p>\n<p>\u6700\u7ec8\u8868\u73b0\u4e3a\uff1a\u5728\u8bad\u7ec3\u6d4b\u8bd5\u96c6\u4e0a\u5206\u6570\u6301\u7eed\u98d9\u5347\uff0c\u4f46\u5207\u6362\u5230\u5168\u65b0\u771f\u5b9e\u573a\u666f\u3001\u672a\u89c1\u8fc7\u7684\u6d4b\u8bd5\u4efb\u52a1\uff08Held-out Set\uff09\u65f6\uff0c\u80fd\u529b\u5927\u5e45\u7f29\u6c34\uff0c\u6cdb\u5316\u6027\u6781\u5dee\u3002<\/p>\n<h4>2.\u00a0\u8bc4\u4f30\u566a\u58f0\u8bef\u5bfc\u8fed\u4ee3\u65b9\u5411<\/h4>\n<p>\u667a\u80fd\u4f53\u4efb\u52a1\u8bc4\u4f30\u5b58\u5728\u5929\u7136\u968f\u673a\u6027\uff0c\u5355\u6b21\u6d4b\u8bd5\u7684\u5206\u6570\u6ce2\u52a8\u53ef\u80fd\u662f\u73af\u5883\u3001\u91c7\u6837\u6982\u7387\u5bfc\u81f4\u7684\u5076\u7136\u73b0\u8c61\uff0c\u800c\u975e\u80fd\u529b\u771f\u5b9e\u63d0\u5347\u3002<\/p>\n<p>\u4f20\u7edf\u65b9\u6848\u65e0\u6cd5\u533a\u5206\u00a0\u201c\u771f\u5b9e\u80fd\u529b\u8fdb\u6b65\u201d\u00a0\u548c\u00a0\u201c\u968f\u673a\u566a\u58f0\u6da8\u5206\u201d\uff0c\u7ecf\u5e38\u57fa\u4e8e\u865a\u5047\u63d0\u5347\u8fed\u4ee3\u4f18\u5316\uff0c\u5bfc\u81f4\u00a0Agent\u00a0\u8fed\u4ee3\u8dd1\u504f\u3001\u80fd\u529b\u9000\u5316\u3002<\/p>\n<h4>3.\u00a0\u7b97\u529b\u5197\u4f59\u4e0e\u590d\u6742\u5ea6\u7d2f\u79ef<\/h4>\n<p>\u65e0\u7ea6\u675f\u8fed\u4ee3\u4e2d\uff0cAgent\u00a0\u4f1a\u901a\u8fc7\u5806\u53e0\u8d85\u957f\u63d0\u793a\u8bcd\u3001\u5197\u4f59\u5de5\u5177\u8c03\u7528\u3001\u590d\u6742\u4e0a\u4e0b\u6587\u903b\u8f91\u6362\u53d6\u5fae\u5c0f\u5206\u6570\u63d0\u5347\uff0c\u5bfc\u81f4\u5355\u6b21\u4efb\u52a1\u00a0Token\u00a0\u6d88\u8017\u6301\u7eed\u66b4\u6da8\u3002<\/p>\n<p>\u8fed\u4ee3\u8f6e\u6b21\u8d8a\u591a\uff0c\u6846\u67b6\u590d\u6742\u5ea6\u8d8a\u9ad8\u3001\u63a8\u7406\u6210\u672c\u8d8a\u5927\uff0c\u6700\u7ec8\u5f62\u6210\u00a0\u201c\u9ad8\u5206\u4f4e\u80fd\u3001\u9ad8\u6210\u672c\u4f4e\u6536\u76ca\u201d\u00a0\u7684\u7578\u5f62\u72b6\u6001\u3002<\/p>\n<h4>4.\u00a0\u6f0f\u6d1e\u8fed\u4ee3\u4e0e\u9690\u6027\u98ce\u9669\u7d2f\u79ef<\/h4>\n<p>\u666e\u901a\u8fed\u4ee3\u65e0\u524d\u7f6e\u7b5b\u67e5\u673a\u5236\uff0cAgent\u00a0\u53ef\u80fd\u751f\u6210\u6f0f\u6d1e\u914d\u7f6e\u3001\u7b54\u6848\u8bb0\u5fc6\u3001\u5bc6\u94a5\u6cc4\u9732\u3001\u786c\u7f16\u7801\u4efb\u52a1\u00a0ID\u00a0\u7b49\u8fdd\u89c4\u4fee\u6539\u3002<\/p>\n<p>\u8fd9\u4e9b\u95ee\u9898\u4e0d\u4f1a\u7acb\u523b\u66b4\u9732\uff0c\u4f1a\u968f\u7740\u8fed\u4ee3\u7d2f\u79ef\uff0c\u6700\u7ec8\u5bfc\u81f4\u667a\u80fd\u4f53\u5d29\u6e83\u3001\u4efb\u52a1\u62a5\u9519\u3001\u5b89\u5168\u98ce\u9669\u6cc4\u9732\uff0c\u4e14\u65e0\u6cd5\u8ffd\u6eaf\u95ee\u9898\u6839\u6e90\u3002<\/p>\n<h4>5.\u00a0\u8fed\u4ee3\u65e0\u5f52\u56e0\u3001\u65e0\u6536\u655b\u6027<\/h4>\n<p>\u4f20\u7edf\u00a0Agent\u00a0\u8fed\u4ee3\u65e0\u5386\u53f2\u8bb0\u5f55\u7ba1\u7406\u3001\u65e0\u526a\u679d\u673a\u5236\uff0c\u65e0\u6548\u4fee\u6539\u3001\u5931\u8d25\u65b9\u6848\u4f1a\u53cd\u590d\u91cd\u8bd5\uff0c\u6210\u529f\u7684\u6709\u6548\u673a\u5236\u65e0\u6cd5\u6c89\u6dc0\u3002<\/p>\n<p>\u8fed\u4ee3\u8fc7\u7a0b\u6742\u4e71\u65e0\u7ae0\uff0c\u65e2\u65e0\u6cd5\u6536\u655b\u5230\u6700\u4f18\u6846\u67b6\uff0c\u4e5f\u65e0\u6cd5\u5b9a\u4f4d\u6709\u6548\u4f18\u5316\u70b9\uff0c\u957f\u671f\u8fed\u4ee3\u4e0d\u4ec5\u65e0\u6cd5\u8fdb\u6b65\uff0c\u53cd\u800c\u4f1a\u5bfc\u81f4\u6846\u67b6\u9000\u5316\u3002<\/p>\n<h2>\u8c37\u6b4c RRSI \u63d0\u4f9b\u7684\u9488\u5bf9\u6027\u89e3\u51b3\u65b9\u6848<\/h2>\n<p>\u9488\u5bf9\u4e0a\u8ff0\u4e94\u5927\u6280\u672f\u75db\u70b9\uff0c\u8c37\u6b4c\u00a0RRSI\u00a0\u4ece<strong>\u8bc4\u4f30\u6821\u51c6\u3001\u8fed\u4ee3\u7b5b\u9009\u3001\u9884\u7b97\u7ea6\u675f\u3001\u98ce\u9669\u7b5b\u67e5\u3001\u5386\u53f2\u7ba1\u63a7<\/strong>\u4e94\u5927\u7ef4\u5ea6\uff0c\u642d\u5efa\u4e86\u4e00\u5957\u5b8c\u6574\u7684\u6b63\u5219\u5316\u89e3\u51b3\u65b9\u6848\uff0c\u6240\u6709\u6838\u5fc3\u903b\u8f91\u5747\u4e3a\u53ef\u843d\u5730\u7684\u00a0Python\u00a0\u4ee3\u7801\u5b9e\u73b0\uff0c\u65e0\u00a0GPU\u3001\u65e0\u6570\u636e\u96c6\u3001\u65e0\u00a0API\u00a0\u5bc6\u94a5\u5373\u53ef\u79bb\u7ebf\u8c03\u8bd5\u3002<\/p>\n<h4>1.\u00a0\u566a\u58f0\u6821\u51c6\u673a\u5236\uff1a\u533a\u5206\u771f\u5b9e\u63d0\u5347\u4e0e\u968f\u673a\u6ce2\u52a8<\/h4>\n<p>RRSI\u00a0\u6452\u5f03\u56fa\u5b9a\u9608\u503c\u5224\u65ad\uff0c\u901a\u8fc7\u00a0calibrate\u00a0\u6821\u51c6\u51fd\u6570\u5bf9\u540c\u4e00\u667a\u80fd\u4f53\u6846\u67b6\u8fdb\u884c\u591a\u6b21\u91cd\u590d\u8bc4\u4f30\uff0c\u8ba1\u7b97\u5206\u6570\u6ce2\u52a8\u6807\u51c6\u5dee\uff0c\u751f\u6210\u52a8\u6001\u566a\u58f0\u5e26\u00a0delta\u3002\u4efb\u4f55\u5206\u6570\u63d0\u5347\u5c0f\u4e8e\u00a0delta\u00a0\u7684\u4fee\u6539\uff0c\u5747\u5224\u5b9a\u4e3a\u968f\u673a\u566a\u58f0\uff0c\u76f4\u63a5\u62d2\u7edd\uff0c\u907f\u514d\u8bef\u5bfc\u8fed\u4ee3\u3002<\/p>\n<p><strong>\u5b9e\u6d4b\u6570\u636e\u663e\u793a<\/strong>\uff1a\u968f\u7740\u8bc4\u4f30\u6b21\u6570\u63d0\u5347\uff0cdelta\u00a0\u503c\u53ef\u4ece\u00a00.096\u00a0\u6536\u655b\u81f3\u00a00.025\uff0c\u4e0e\u5b98\u65b9\u8bba\u6587\u00a00.004-0.020\u00a0\u7684\u6807\u51c6\u533a\u95f4\u5b8c\u5168\u5339\u914d\uff0c\u8bc4\u4f30\u7cbe\u5ea6\u8fdc\u8d85\u4f20\u7edf\u65b9\u6848\u3002<\/p>\n<h4>2.\u00a0\u53cc\u5c42\u7b5b\u9009\u7b97\u6cd5\uff08Algorithm 2\uff09\uff1a\u675c\u7edd\u8d2a\u5fc3\u5237\u5206<\/h4>\n<p>RRSI\u00a0\u6838\u5fc3\u9009\u62e9\u7b97\u6cd5\u8bbe\u7f6e\u4e09\u91cd\u521a\u6027\u89c4\u5219\uff0c\u6539\u5199\u4e86\u4f20\u7edf\u00a0\u201c\u9ad8\u5206\u5373\u6700\u4f18\u201d\u00a0\u7684\u7b5b\u9009\u903b\u8f91\uff1a<\/p>\n<p>\u2022\u00a0<strong>\u4e0b\u9650\u4fdd\u62a4\u89c4\u5219<\/strong>\uff1a\u5019\u9009\u65b9\u6848\u5206\u6570\u4e0d\u5f97\u4f4e\u4e8e\u5386\u53f2\u6700\u9ad8\u5206-\u566a\u58f0\u5e26\uff0c\u675c\u7edd\u6574\u4f53\u80fd\u529b\u9000\u5316<\/p>\n<p>\u2022\u00a0<strong>\u7b97\u529b\u6210\u672c\u89c4\u5219<\/strong>\uff1a\u5206\u6570\u63d0\u5347\u5fc5\u987b\u5339\u914d\u00a0Token\u00a0\u6d88\u8017\u589e\u91cf\uff0c\u516c\u5f0f\uff1adC \u2264 0.10 + 40*dS\uff0c\u675c\u7edd\u9ad8\u6210\u672c\u4f4e\u6536\u76ca\u4f18\u5316<\/p>\n<p>\u2022\u00a0<strong>\u7ed3\u6784\u5316\u62e9\u4f18\u89c4\u5219<\/strong>\uff1a\u566a\u58f0\u5e26\u5185\u540c\u5206\u65b9\u6848\uff0c\u4f18\u5148\u9009\u62e9\u4f4e\u6210\u672c\u3001\u65b0\u7ed3\u6784\u7684\u4fee\u6539\uff0c\u9f13\u52b1\u521b\u65b0\u4f18\u5316\u3001\u7cbe\u7b80\u6846\u67b6<\/p>\n<p>\u8be5\u89c4\u5219\u53ef\u5b9e\u73b0\u53cd\u76f4\u89c9\u4f18\u8d28\u8fed\u4ee3\uff1a<strong>\u90e8\u5206\u5c0f\u5e45\u964d\u5206\u4f46\u5927\u5e45\u964d\u672c\u7684\u65b9\u6848\uff0c\u4f1a\u88ab\u4fdd\u7559\uff0c\u5b9e\u73b0\u667a\u80fd\u4f53\u8f7b\u91cf\u5316\u4f18\u5316\u3002<\/strong><\/p>\n<h4>3.\u00a0\u9000\u706b\u7f16\u8f91\u9884\u7b97\uff1a\u53ef\u63a7\u8fed\u4ee3\u3001\u7cbe\u51c6\u5f52\u56e0<\/h4>\n<p>\u901a\u8fc7\u00a0edit_budget\u00a0\u51fd\u6570\u5b9e\u73b0\u4f59\u5f26\u9000\u706b\u8fed\u4ee3\u9884\u7b97\uff0c\u8fed\u4ee3\u524d\u671f\u5141\u8bb8\u6700\u591a\u00a04\u00a0\u5904\u534f\u540c\u4fee\u6539\uff0c\u9010\u6b65\u9012\u51cf\u81f3\u00a02\u00a0\u5904\uff0c\u9650\u5236\u5355\u8f6e\u4fee\u6539\u6570\u91cf\u3002<\/p>\n<p>\u524d\u671f\u5145\u5206\u63a2\u7d22\u4f18\u5316\u7a7a\u95f4\uff0c\u540e\u671f\u7cbe\u7b80\u4fee\u6539\u7c92\u5ea6\uff0c\u8ba9\u6bcf\u4e00\u8f6e\u6709\u6548\u8fed\u4ee3\u90fd\u53ef\u7cbe\u51c6\u5f52\u56e0\u5230\u5177\u4f53\u7ec4\u4ef6\uff0c\u907f\u514d\u590d\u6742\u4fee\u6539\u53e0\u52a0\u5bfc\u81f4\u7684\u4e0d\u53ef\u63a7\u95ee\u9898\u3002<\/p>\n<h4>4.\u00a0\u53cc\u5c42\u5ba1\u67e5\u673a\u5236\uff1a\u62e6\u622a\u6240\u6709\u9690\u6027\u98ce\u9669<\/h4>\n<p>RRSI\u00a0\u642d\u5efa<strong>\u786e\u5b9a\u6027\u9884\u68c0\u67e5+\u5927\u6a21\u578b\u610f\u56fe\u590d\u6838<\/strong>\u53cc\u5c42\u7b5b\u67e5\u4f53\u7cfb\uff1a<\/p>\n<p>\u2022\u00a0<strong>\u786e\u5b9a\u6027\u9884\u68c0\u67e5<\/strong>\uff1a\u901a\u8fc7\u6b63\u5219\u5339\u914d\uff0c\u76f4\u63a5\u62e6\u622a\u8bb0\u5fc6\u4efb\u52a1\u7b54\u6848\u3001\u8bfb\u53d6\u8bc4\u5206\u89c4\u5219\u3001\u6cc4\u9732\u5bc6\u94a5\u3001\u786c\u7f16\u7801\u4efb\u52a1\u00a0ID\u00a0\u7b49\u8fdd\u89c4\u4fee\u6539\uff0c\u65e0\u9700\u8c03\u7528\u5927\u6a21\u578b\uff0c\u96f6\u6210\u672c\u98ce\u63a7<\/p>\n<p>\u2022\u00a0<strong>LLM \u610f\u56fe\u590d\u6838<\/strong>\uff1a\u5e72\u51c0\u5408\u89c4\u7684\u4fee\u6539\u4ea4\u7531\u00a0Claude Opus\u00a0\u590d\u6838\uff0c\u6392\u67e5\u9690\u6027\u903b\u8f91\u6f0f\u6d1e\u3001\u8de8\u573a\u666f\u9002\u914d\u95ee\u9898<\/p>\n<h4>5.\u00a0\u5386\u53f2\u8bb0\u5f55\u4e0e\u526a\u679d\u673a\u5236\uff1a\u5b9e\u73b0\u8fed\u4ee3\u6536\u655b<\/h4>\n<p>\u901a\u8fc7\u00a0History\u00a0\u6a21\u5757\u5b8c\u6574\u8bb0\u5f55\u6bcf\u4e00\u8f6e\u4fee\u6539\u7684\u7ec4\u4ef6\u3001\u5206\u6570\u53d8\u5316\u3001\u7b97\u529b\u6210\u672c\u3001\u843d\u5730\u6548\u679c\uff0c\u81ea\u52a8\u7edf\u8ba1\u5404\u7ec4\u4ef6\u5386\u53f2\u6536\u76ca\u3002<\/p>\n<p>\u5bf9\u957f\u671f\u65e0\u6b63\u5411\u63d0\u5347\u3001\u5197\u4f59\u65e0\u6548\u7684\u6846\u67b6\u7ec4\u4ef6\u8fdb\u884c\u81ea\u52a8\u526a\u679d\uff1b\u8fed\u4ee3\u505c\u6ede\u65f6\uff0c\u81ea\u52a8\u5f15\u5bfc\u667a\u80fd\u4f53\u63a2\u7d22\u672a\u5c1d\u8bd5\u8fc7\u7684\u4f18\u5316\u65b9\u5411\uff0c\u6253\u7834\u5c40\u90e8\u6700\u4f18\u50f5\u5c40\u3002<\/p>\n<h4>6.\u00a0\u521d\u59cb\u5316\u4ee3\u7801\uff08\u53ef\u76f4\u63a5\u8fd0\u884c\uff09<\/h4>\n<p>\u4ee5\u4e0b\u4e3a\u00a0RRSI\u00a0\u5b98\u65b9\u4ed3\u5e93\u90e8\u7f72\u6838\u5fc3\u4ee3\u7801\uff0c\u9501\u5b9a\u7a33\u5b9a\u7248\u672c\uff0c\u53ef\u76f4\u63a5\u5b8c\u6210\u73af\u5883\u521d\u59cb\u5316\u4e0e\u6838\u5fc3\u6a21\u5757\u6620\u5c04\uff1a<\/p>\n<div>\n<div class=\"hljs-cto\">\n<div class=\"hljs-cto\">\n<p><button class=\"copy_btn disable\" data-clipboard-target=\"#code_id_0\">\u590d\u5236<\/button><\/p>\n<div class=\"code-toolbar\">\n<pre class=\"has-pre-numbering language-plain\" tabindex=\"0\"><code class=\"language-plain\">Python\r\nimport os\r\nimport sys\r\nimport json\r\nimport math\r\nimport copy\r\nimport random\r\nimport tempfile\r\nimport textwrap\r\nimport traceback\r\nimport subprocess\r\nimport statistics as st\r\nfrom pathlib import Path\r\n\r\n\r\n# \u5b89\u88c5\u6307\u5b9a\u7248\u672cRRSI\u4f9d\u8d56\uff08\u65e0PyPI\u53d1\u5e03\uff0c\u76f4\u63a5\u62c9\u53d6\u5b98\u65b9\u4ed3\u5e93\uff09\r\nsubprocess.run([sys.executable, \"-m\", \"pip\", \"install\", \"-q\", \"git+https:\/\/github.com\/google-research\/rrsi.git@be50316e1db05914068a973f322770ef08ed7ba1\"], check=True)\r\n\r\n\r\n# \u5bfc\u5165RRSI\u6838\u5fc3\u6a21\u5757\r\nfrom importlib.metadata import version\r\nfrom rrsi.config import RRSIConfig\r\nfrom rrsi.evaluate import TaskResult, EvalResult, aggregate, evaluate\r\nfrom rrsi.calibrate import calibrate\r\nfrom rrsi.selection import Candidate, cost_rule, judge, select_round\r\nfrom rrsi.schedule import edit_budget, budget_table\r\nfrom rrsi.history import History, stall_flag, exploration\r\nfrom rrsi.components import K, K_STR, normalize, novelty\r\nfrom rrsi.critic import precheck, review\r\nfrom rrsi.domain import Domain\r\n\r\n\r\n# \u6253\u5370\u73af\u5883\u4e0e\u6838\u5fc3\u6a21\u5757\u6620\u5c04\u5173\u7cfb\r\nprint(f\"  rrsi {version('rrsi')}  |  anthropic {version('anthropic')}  |  Python {sys.version.split()[0]}\")\r\nCFG = RRSIConfig()\r\nprint(f\"\\n  \u5b98\u65b9\u9ed8\u8ba4\u8d85\u53c2: T={CFG.T}\u8f6e\u8fed\u4ee3, k={CFG.k}\u6b21\u4efb\u52a1\u5c1d\u8bd5, m={CFG.m}\u4e2a\u5019\u9009\u65b9\u6848\/\u8f6e\")<\/code><\/pre>\n<ul id=\"code_id_0\" class=\"pre-numbering\">\n<li>1.<\/li>\n<li>2.<\/li>\n<li>3.<\/li>\n<li>4.<\/li>\n<li>5.<\/li>\n<li>6.<\/li>\n<li>7.<\/li>\n<li>8.<\/li>\n<li>9.<\/li>\n<li>10.<\/li>\n<li>11.<\/li>\n<li>12.<\/li>\n<li>13.<\/li>\n<li>14.<\/li>\n<li>15.<\/li>\n<li>16.<\/li>\n<li>17.<\/li>\n<li>18.<\/li>\n<li>19.<\/li>\n<li>20.<\/li>\n<li>21.<\/li>\n<li>22.<\/li>\n<li>23.<\/li>\n<li>24.<\/li>\n<li>25.<\/li>\n<li>26.<\/li>\n<li>27.<\/li>\n<li>28.<\/li>\n<li>29.<\/li>\n<li>30.<\/li>\n<li>31.<\/li>\n<li>32.<\/li>\n<li>33.<\/li>\n<li>34.<\/li>\n<li>35.<\/li>\n<li>36.<\/li>\n<\/ul>\n<div class=\"toolbar\"><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<h2>\u5b9e\u6218\u6848\u4f8b\u5206\u4eab\uff1aRRSI vs\u00a0\u4f20\u7edf\u8d2a\u5fc3\u8fed\u4ee3\u6548\u679c\u5bf9\u6bd4<\/h2>\n<p>\u4e3a\u76f4\u89c2\u9a8c\u8bc1\u00a0RRSI\u00a0\u7684\u4f18\u5316\u6548\u679c\uff0c\u6211\u4eec\u57fa\u4e8e\u8c37\u6b4c\u5b98\u65b9\u6a21\u62df\u667a\u80fd\u4f53\u73af\u5883\uff0c\u642d\u5efa<strong>\u8d2a\u5fc3\u8fed\u4ee3\u3001\u4ec5\u98ce\u63a7\u8fed\u4ee3\u3001RRSI\u00a0\u5b8c\u6574\u8fed\u4ee3<\/strong>\u4e09\u7ec4\u5bf9\u7167\u5b9e\u9a8c\uff0c\u8986\u76d6\u00a040\u00a0\u4e2a\u6f14\u5316\u4efb\u52a1\u300180\u00a0\u4e2a\u672a\u77e5\u6d4b\u8bd5\u4efb\u52a1\uff0c\u901a\u8fc7\u00a08\u00a0\u7ec4\u968f\u673a\u79cd\u5b50\u91cd\u590d\u5b9e\u9a8c\uff0c\u6570\u636e\u771f\u5b9e\u53ef\u590d\u73b0\u3002<\/p>\n<h4>1.\u00a0\u5b9e\u9a8c\u6838\u5fc3\u914d\u7f6e<\/h4>\n<p>\u2022\u00a0<strong>\u4efb\u52a1\u7c7b\u578b<\/strong>\uff1a\u89e3\u6790\u3001\u641c\u7d22\u3001\u7f16\u8f91\u3001\u6d4b\u8bd5\u56db\u5927\u7c7b\u667a\u80fd\u4f53\u901a\u7528\u4efb\u52a1<\/p>\n<p>\u2022\u00a0<strong>\u8fed\u4ee3\u8f6e\u6b21<\/strong>\uff1a20\u00a0\u8f6e\u6807\u51c6\u8fed\u4ee3<\/p>\n<p>\u2022\u00a0<strong>\u8bc4\u4f30\u65b9\u5f0f<\/strong>\uff1a\u6bcf\u4efb\u52a1\u00a02\u00a0\u6b21\u91cd\u590d\u6d4b\u8bd5\uff0c\u6821\u51c6\u8bc4\u4f30\u566a\u58f0<\/p>\n<p>\u2022\u00a0<strong>\u5bf9\u6bd4\u7ef4\u5ea6<\/strong>\uff1a\u6f14\u5316\u96c6\u5206\u6570\u3001\u672a\u77e5\u4efb\u52a1\u6cdb\u5316\u5206\u6570\u3001Token\u00a0\u7b97\u529b\u6210\u672c\u3001\u7b54\u6848\u8bb0\u5fc6\u7387\u3001\u65e0\u6548\u8bc4\u4f30\u6b21\u6570<\/p>\n<h4>2.\u00a0\u6838\u5fc3\u5b9e\u9a8c\u6570\u636e\u5bf9\u6bd4<\/h4>\n<table class=\"data-table\" data-transient-attributes=\"class\" data-width=\"688px\">\n<colgroup data-id=\"c7104f7d-bRBD4RCm\">\n<col span=\"1\" width=\"134.2\" data-id=\"cd89ecb0-eB36g7R2\" \/>\n<col span=\"1\" width=\"134.2\" data-id=\"cd89ecb0-COgdPi00\" \/>\n<col span=\"1\" width=\"151.2\" data-id=\"cd89ecb0-WfhnTShf\" \/>\n<col span=\"1\" width=\"134.2\" data-id=\"cd89ecb0-Wpkk4CUW\" \/>\n<col span=\"1\" width=\"134.2\" data-id=\"cd89ecb0-dEGq4MV1\" \/><\/colgroup>\n<tbody data-id=\"t6d5e859-K0SX5FMD\">\n<tr data-id=\"t31e458f-gUN0K5jB\">\n<td data-id=\"t6267798-hkmbUDQU\" data-transient-attributes=\"table-cell-selection\"><strong>\u8fed\u4ee3\u6a21\u5f0f<\/strong><\/td>\n<td data-id=\"t6267798-mh0L4lfb\" data-transient-attributes=\"table-cell-selection\"><strong>\u6f14\u5316\u96c6\u5206\u6570<\/strong><\/td>\n<td data-id=\"t6267798-0bXJgBEm\" data-transient-attributes=\"table-cell-selection\"><strong>\u672a\u77e5\u4efb\u52a1\u6cdb\u5316\u5206\u6570<\/strong><\/td>\n<td data-id=\"t6267798-PnS5iggD\" data-transient-attributes=\"table-cell-selection\"><strong>Token\u00a0\u76f8\u5bf9\u6210\u672c<\/strong><\/td>\n<td class=\"table-last-row\" data-id=\"t6267798-ZDSfZpTP\" data-transient-attributes=\"table-cell-selection\"><strong>\u7b54\u6848\u8bb0\u5fc6\u6570\u91cf<\/strong><\/td>\n<\/tr>\n<tr data-id=\"t31e458f-CgSeEFRT\">\n<td data-id=\"t6267798-jddFUS9V\" data-transient-attributes=\"table-cell-selection\">\u4f20\u7edf\u8d2a\u5fc3\u8fed\u4ee3<\/td>\n<td data-id=\"t6267798-SIJhiega\" data-transient-attributes=\"table-cell-selection\">0.681\uff08\u6700\u9ad8\uff09<\/td>\n<td data-id=\"t6267798-mqgZ3MLO\" data-transient-attributes=\"table-cell-selection\">0.523\uff08\u6700\u4f4e\uff09<\/td>\n<td data-id=\"t6267798-St4cEQ5m\" data-transient-attributes=\"table-cell-selection\">2.97x\uff08\u6781\u9ad8\uff09<\/td>\n<td class=\"table-last-row\" data-id=\"t6267798-515OIR56\" data-transient-attributes=\"table-cell-selection\">10.1\uff08\u4e25\u91cd\u8fc7\u62df\u5408\uff09<\/td>\n<\/tr>\n<tr data-id=\"t31e458f-J3hTXW0b\">\n<td data-id=\"t6267798-dE2uvZ2o\" data-transient-attributes=\"table-cell-selection\">\u4ec5\u98ce\u63a7\u8fed\u4ee3<\/td>\n<td data-id=\"t6267798-xYEw5862\" data-transient-attributes=\"table-cell-selection\">0.652\uff08\u4e2d\u7b49\uff09<\/td>\n<td data-id=\"t6267798-vCTi4EfF\" data-transient-attributes=\"table-cell-selection\">0.587\uff08\u4e2d\u7b49\uff09<\/td>\n<td data-id=\"t6267798-nOZchK5d\" data-transient-attributes=\"table-cell-selection\">2.14x\uff08\u4e2d\u7b49\uff09<\/td>\n<td class=\"table-last-row\" data-id=\"t6267798-nhNjZhDl\" data-transient-attributes=\"table-cell-selection\">0.3\uff08\u57fa\u672c\u65e0\u8bb0\u5fc6\uff09<\/td>\n<\/tr>\n<tr data-id=\"t31e458f-NmlPJPQ3\">\n<td class=\"table-last-column\" data-id=\"t6267798-L35qFAr5\" data-transient-attributes=\"table-cell-selection\">RRSI\u00a0\u6b63\u5219\u5316\u8fed\u4ee3<\/td>\n<td class=\"table-last-column\" data-id=\"t6267798-Wk5fRMbR\" data-transient-attributes=\"table-cell-selection\">0.616\uff08\u53ef\u63a7\u5c0f\u5e45\u4e0b\u964d\uff09<\/td>\n<td class=\"table-last-column\" data-id=\"t6267798-ae6FbqQl\" data-transient-attributes=\"table-cell-selection\">0.616\uff08\u6700\u9ad8\u6cdb\u5316\uff09<\/td>\n<td class=\"table-last-column\" data-id=\"t6267798-KjNefvpQ\" data-transient-attributes=\"table-cell-selection\">1.53x\uff08\u6700\u4f4e\u6210\u672c\uff09<\/td>\n<td class=\"table-last-column table-last-row\" data-id=\"t6267798-dipp4YfG\" data-transient-attributes=\"table-cell-selection\">0\uff08\u96f6\u8fc7\u62df\u5408\uff09<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4>3.\u00a0\u6848\u4f8b\u7ed3\u679c<\/h4>\n<p>\u2022\u00a0<strong>\u7834\u9664\u9ad8\u5206\u9677\u9631<\/strong>\uff1a\u4f20\u7edf\u8d2a\u5fc3\u8fed\u4ee3\u770b\u4f3c\u8bad\u7ec3\u5206\u6570\u6700\u9ad8\uff0c\u5b9e\u5219\u4e25\u91cd\u8fc7\u62df\u5408\uff0c\u672a\u77e5\u573a\u666f\u80fd\u529b\u5927\u5e45\u8870\u51cf\uff1bRRSI\u00a0\u5c0f\u5e45\u727a\u7272\u8bad\u7ec3\u96c6\u5206\u6570\uff0c\u6362\u6765\u663e\u8457\u7684\u6cdb\u5316\u80fd\u529b\u63d0\u5347\uff0c\u9002\u914d\u771f\u5b9e\u843d\u5730\u573a\u666f\u3002<\/p>\n<p>\u2022\u00a0<strong>\u6781\u81f4\u964d\u672c\u589e\u6548<\/strong>\uff1a\u540c\u7b49\u4efb\u52a1\u6548\u679c\u4e0b\uff0cRRSI\u00a0\u7b97\u529b\u6210\u672c\u4ec5\u4e3a\u4f20\u7edf\u65b9\u6848\u7684\u00a051%\uff0c\u4e14\u968f\u7740\u8bc4\u4f30\u7cbe\u5ea6\u63d0\u5347\uff0c\u5dee\u8ddd\u6301\u7eed\u62c9\u5927\uff0c\u9ad8\u7cbe\u5ea6\u573a\u666f\u4e0b\u4f20\u7edf\u65b9\u6848\u7b97\u529b\u6d88\u8017\u53ef\u8fbe\u00a0RRSI\u00a0\u7684\u00a06.5\u500d\u3002<\/p>\n<p>\u2022\u00a0<strong>\u5f7b\u5e95\u675c\u7edd\u5237\u5206<\/strong>\uff1aRRSI\u00a0\u98ce\u63a7\u673a\u5236\u53ef\u63d0\u524d\u62e6\u622a\u00a099%\u00a0\u7684\u7b54\u6848\u8bb0\u5fc6\u3001\u89c4\u5219\u7a83\u53d6\u7b49\u8fdd\u89c4\u4f18\u5316\uff0c\u4ece\u6839\u6e90\u89e3\u51b3\u667a\u80fd\u4f53\u00a0\u201c\u5e94\u8bd5\u5316\u201d\u00a0\u95ee\u9898\u3002<\/p>\n<p>\u2022<strong>\u00a0\u8fed\u4ee3\u7a33\u5b9a\u6027\u62c9\u6ee1<\/strong>\uff1a30\u00a0\u8f6e\u957f\u5468\u671f\u8fed\u4ee3\u4e2d\uff0cRRSI\u00a0\u4ec5\u4fdd\u7559\u00a010\u00a0\u8f6e\u6709\u6548\u4f18\u5316\uff0c\u62d2\u7edd\u6240\u6709\u566a\u58f0\u6ce2\u52a8\u3001\u65e0\u6548\u5806\u53e0\u3001\u9ad8\u6210\u672c\u4fee\u6539\uff0c\u8fed\u4ee3\u8fc7\u7a0b\u53ef\u63a7\u3001\u53ef\u8ffd\u6eaf\u3001\u53ef\u6536\u655b\u3002<\/p>\n<h4>4.\u00a0\u771f\u5b9e\u4e1a\u52a1\u573a\u666f\u843d\u5730\u6548\u679c<\/h4>\n<p>\u8c37\u6b4c\u5b98\u65b9\u5728\u4ee3\u7801\u3001\u5de5\u4f5c\u7a7a\u95f4\u3001\u5de5\u7a0b\u8bbe\u8ba1\u7684\u4e09\u5927\u771f\u5b9e\u573a\u666f\u9a8c\u8bc1\uff1a<\/p>\n<p><strong>RRSI\u00a0\u53ef\u5b9e\u73b0\u8de8\u57fa\u51c6\u6cdb\u5316\u80fd\u529b\u5168\u9762\u63d0\u5347\uff0cTerminal-Bench\u00a0\u4ee3\u7801\u4efb\u52a1\u8fc1\u79fb\u81f3\u00a0SWE-bench Verified\u00a0\u5206\u6570\u63d0\u5347\u00a01.8\u00a0\u5206\uff0c\u5de5\u7a0b\u8bbe\u8ba1\u4efb\u52a1\u00a0Frontier-Eng\u00a0\u5206\u6570\u63d0\u5347\u00a04.3\u00a0\u5206\u3002<\/strong><\/p>\n<p>\u6240\u6709\u672a\u77e5\u6d4b\u8bd5\u96c6\u5747\u5b9e\u73b0\u6b63\u5411\u6536\u76ca\uff0c\u4e14\u6846\u67b6\u590d\u6742\u5ea6\u6301\u7eed\u4f18\u5316\u3002<\/p>\n<h2>\u603b\u7ed3\uff1aRRSI\u00a0\u91cd\u65b0\u5b9a\u4e49\u00a0AI\u00a0\u667a\u80fd\u4f53\u81ea\u8fdb\u5316<\/h2>\n<p>RRSI\u00a0\u7684\u6838\u5fc3\u4ef7\u503c\uff0c\u4ece\u6765\u4e0d\u662f\u8ba9\u00a0AI\u00a0\u667a\u80fd\u4f53\u00a0\u201c\u66f4\u5feb\u6539\u81ea\u5df1\u201d\uff0c\u800c\u662f\u8ba9\u667a\u80fd\u4f53\u00a0\u201c\u66f4\u806a\u660e\u3001\u66f4\u7a33\u5b9a\u3001\u66f4\u7701\u94b1\u5730\u8fed\u4ee3\u81ea\u5df1\u201d\u3002<\/p>\n<p>\u5b83\u8df3\u51fa\u4e86\u4f20\u7edf\u00a0AI\u00a0\u81ea\u8fdb\u5316\u00a0\u201c\u552f\u5206\u6570\u8bba\u201d\u00a0\u7684\u8bef\u533a\uff0c\u7528\u6b63\u5219\u5316\u7ea6\u675f\u89e3\u51b3\u4e86\u957f\u671f\u56f0\u6270\u884c\u4e1a\u7684\u8fc7\u62df\u5408\u3001\u566a\u58f0\u8bef\u5bfc\u3001\u7b97\u529b\u6d6a\u8d39\u3001\u98ce\u9669\u7d2f\u79ef\u56db\u5927\u96be\u9898\u3002<\/p>\n<p><strong>\u5bf9\u4e8e\u5f00\u53d1\u8005\u800c\u8a00\uff0cRRSI\u00a0\u95e8\u69db\u4f4e\u3001\u843d\u5730\u6027\u5f3a<\/strong>\uff1a\u65e0\u9700\u5fae\u8c03\u5927\u6a21\u578b\u3001\u65e0\u9700\u6d77\u91cf\u6570\u636e\u3001\u65e0\u9700\u9ad8\u7aef\u7b97\u529b\uff0c\u4ec5\u901a\u8fc7\u8f7b\u91cf\u5316\u6846\u67b6\u8fed\u4ee3\uff0c\u5c31\u80fd\u5927\u5e45\u63d0\u5347\u667a\u80fd\u4f53\u7684\u901a\u7528\u6027\u548c\u7a33\u5b9a\u6027\uff0c\u662f\u5f53\u524d\u00a0Agent\u00a0\u5de5\u7a0b\u843d\u5730\u3001\u6301\u7eed\u8fed\u4ee3\u7684\u6700\u4f18\u65b9\u6848\u4e4b\u4e00\u3002<\/p>\n<p>\u672a\u6765\u00a0AI\u00a0\u667a\u80fd\u4f53\u7684\u7ade\u4e89\uff0c\u65e9\u5df2\u4e0d\u662f\u5355\u6b21\u4efb\u52a1\u6548\u679c\u7684\u6bd4\u62fc\uff0c\u800c\u662f<strong>\u53ef\u6301\u7eed\u81ea\u6211\u8fdb\u5316\u80fd\u529b<\/strong>\u7684\u6bd4\u62fc\u3002RRSI\u00a0\u6240\u4ee3\u8868\u7684\u6b63\u5219\u5316\u81ea\u8fdb\u5316\u601d\u8def\uff0c\u5c06\u6210\u4e3a\u4e0b\u4e00\u4ee3\u901a\u7528\u667a\u80fd\u4f53\u7684\u5e95\u5c42\u6280\u672f\u3002<\/p>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_30351\" class=\"pvc_stats total_only  \" data-element-id=\"30351\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/www.ixpower.cn\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"excerpt":{"rendered":"<p>RRSI \u7684\u5168\u79f0\u662f\u00a0Regularized Recursive Self-Improvement of Age [&hellip;]<\/p>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_30351\" class=\"pvc_stats total_only  \" data-element-id=\"30351\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/www.ixpower.cn\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"author":56,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"publish_to_discourse":"","publish_post_category":"","wpdc_auto_publish_overridden":"","wpdc_topic_tags":"","wpdc_pin_topic":"","wpdc_pin_until":"","discourse_post_id":"","discourse_permalink":"","wpdc_publishing_response":"","wpdc_publishing_error":"","footnotes":""},"categories":[3,23,20,80],"tags":[2319],"class_list":["post-30351","post","type-post","status-publish","format-standard","hentry","category-ai","category-23","category-20","category-80","tag--agent--rrsi-ai-"],"_links":{"self":[{"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/posts\/30351","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/comments?post=30351"}],"version-history":[{"count":2,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/posts\/30351\/revisions"}],"predecessor-version":[{"id":30404,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/posts\/30351\/revisions\/30404"}],"wp:attachment":[{"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/media?parent=30351"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/categories?post=30351"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ixpower.cn\/index.php\/wp-json\/wp\/v2\/tags?post=30351"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}