{"id":8099,"date":"2026-09-17T15:00:47","date_gmt":"2026-09-17T07:00:47","guid":{"rendered":"https:\/\/1ow.top\/index.php\/2026\/09\/17\/%e7%bd%97%e7%a6%8f%e8%8e%89%e6%b2%89%e5%af%82%e5%8d%8a%e5%b9%b4%e5%ae%98%e5%ae%a3%e5%b0%8f%e7%b1%b3%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0%ef%bc%81%e7%9b%b4%e6%92%ad%e6%96%b0%e6%a8%a1%e5%9e%8b%e8%ae%ad\/"},"modified":"2026-09-17T15:00:47","modified_gmt":"2026-09-17T07:00:47","slug":"%e7%bd%97%e7%a6%8f%e8%8e%89%e6%b2%89%e5%af%82%e5%8d%8a%e5%b9%b4%e5%ae%98%e5%ae%a3%e5%b0%8f%e7%b1%b3%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0%ef%bc%81%e7%9b%b4%e6%92%ad%e6%96%b0%e6%a8%a1%e5%9e%8b%e8%ae%ad","status":"publish","type":"post","link":"https:\/\/1ow.top\/index.php\/2026\/09\/17\/%e7%bd%97%e7%a6%8f%e8%8e%89%e6%b2%89%e5%af%82%e5%8d%8a%e5%b9%b4%e5%ae%98%e5%ae%a3%e5%b0%8f%e7%b1%b3%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0%ef%bc%81%e7%9b%b4%e6%92%ad%e6%96%b0%e6%a8%a1%e5%9e%8b%e8%ae%ad\/","title":{"rendered":"\u7f57\u798f\u8389\u6c89\u5bc2\u534a\u5e74\u5b98\u5ba3\u5c0f\u7c73\u5f3a\u5316\u5b66\u4e60\uff01\u76f4\u64ad\u65b0\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u4e00\u5c0f\u65f6\u70e73\u4e07\u7f8e\u5143"},"content":{"rendered":"<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/a005ae03ae546c0fdec66bfac6d2a54b.gif'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/a005ae03ae546c0fdec66bfac6d2a54b.gif\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"\u7f57\u798f\u8389\u6c89\u5bc2\u534a\u5e74\u5b98\u5ba3\u5c0f\u7c73\u5f3a\u5316\u5b66\u4e60\uff01\u76f4\u64ad\u65b0\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u4e00\u5c0f\u65f6\u70e73\u4e07\u7f8e\u5143\"\/><\/div><\/figure>\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/93e723a5a9ae8384bb2ad0dfd21c8393.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/93e723a5a9ae8384bb2ad0dfd21c8393.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"\u7f57\u798f\u8389\u6c89\u5bc2\u534a\u5e74\u5b98\u5ba3\u5c0f\u7c73\u5f3a\u5316\u5b66\u4e60\uff01\u76f4\u64ad\u65b0\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u4e00\u5c0f\u65f6\u70e73\u4e07\u7f8e\u5143\"\/><\/div><\/figure>\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/3791e15476221169a3b7e58c12f05716.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/3791e15476221169a3b7e58c12f05716.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"\u7f57\u798f\u8389\u6c89\u5bc2\u534a\u5e74\u5b98\u5ba3\u5c0f\u7c73\u5f3a\u5316\u5b66\u4e60\uff01\u76f4\u64ad\u65b0\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u4e00\u5c0f\u65f6\u70e73\u4e07\u7f8e\u5143\"\/><\/div><\/figure>\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/bdfe66fd8bfa0b831a3f5f0e57f27250.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/1ow.top\/wp-content\/uploads\/2026\/09\/bdfe66fd8bfa0b831a3f5f0e57f27250.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"\u7f57\u798f\u8389\u6c89\u5bc2\u534a\u5e74\u5b98\u5ba3\u5c0f\u7c73\u5f3a\u5316\u5b66\u4e60\uff01\u76f4\u64ad\u65b0\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u4e00\u5c0f\u65f6\u70e73\u4e07\u7f8e\u5143\"\/><\/div><\/figure>\n<p>\u5956\u52b1\u66f2\u7ebf\u3001\u663e\u5361\u6545\u969c\u5168\u516c\u5f00<\/p>\n<p>\u4ec0\u4e48\u60c5\u51b5\uff1f\u5c0f\u7c73\u65b0\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\uff0c\u76f4\u63a5\u5f00\u4e86\u73b0\u573a\u76f4\u64ad\u3002<\/p>\n<p>\u70b9\u8fdb\u53bb\u4e00\u770b\uff0c\u8fd9\u54ea\u662f\u8bad\u7ec3\u73b0\u573a\u554a\uff0c\u8fd9\u5c31\u662f\u70e7\u94b1\u73b0\u573a\uff0c\u4e00\u7728\u773c\u5c31\u662f10\u7f8e\u5200\uff0c\u4e00\u5206\u949f\u5c31\u662f4000\u591a\u5143\u4eba\u6c11\u5e01\u51fa\u53bb\u4e86\u3002<\/p>\n<p>\u5f53\u7136\u70e7\u94b1\u7684\u901f\u5ea6\u53ef\u80fd\u4e0d\u662f\u5747\u5300\u7684\uff0c\u4ecepro\u6a21\u578b\u5f00\u59cb\u8bad\u7ec3\u7b97\u8d7736\u5c0f\u65f6\uff0c\u4e24\u6b3e\u6a21\u578b\u5df2\u7ecf\u82b1\u4e86\u8d85\u8fc7108\u4e07\u7f8e\u5143\uff0c\u5e73\u5747\u6bcf\u5c0f\u65f63\u4e07\u7f8e\u5143\u3002<\/p>\n<p>\u8981\u662f\u522b\u4eba\u5e72\u8fd9\u4e8b\u5f97\u558a\u4e00\u53e5\u8bef\u95ef\u5929\u5bb6\uff0c\u4f46\u662f\u5c0f\u7c73\u5e72\u8fd9\u4e8b\u53ea\u80fd\u558a\u4e00\u53e5\u8bef\u95ef\u7c73\u5bb6\u4e86\u3002<\/p>\n<p>\u5728\u8fd9\u4e2a\u9875\u9762\u4e0a\uff0c\u8bad\u7ec3\u82b1\u4e86\u591a\u5c11\u94b1\u3001\u8dd1\u4e86\u591a\u5c11\u6b65\u3001\u5956\u52b1\u66f2\u7ebf\u6da8\u6ca1\u6da8\u3001\u54ea\u4e2a\u8282\u70b9\u7684\u663e\u5361\u51fa\u4e86\u6545\u969c\uff0c\u5168\u90e8\u516c\u5f00\u53ef\u67e5\u3002<\/p>\n<p>\u89c1\u8fc7\u5f00\u653e\u6743\u91cd\u3001\u5f00\u653e\u8bad\u7ec3\u4ee3\u7801\u548c\u5f00\u653e\u8bad\u7ec3\u6570\u636e\u7684\uff0c\u5f00\u653e\u8bad\u7ec3\u8fc7\u7a0b\u7684\u8fd8\u771f\u662f\u6ca1\u89c1\u8fc7\uff0c\u56f4\u89c2\u7fa4\u4f17\u7eb7\u7eb7\u5927\u547c\u8fc7\u763e\u3002<\/p>\n<p>\u90a3\u4e48\u76ee\u524dMiMo-V2.6\u7684Flash\u548cPro\u4e24\u6b3e\u6a21\u578b\u8bad\u7ec3\u5f97\u600e\u4e48\u6837\u4e86\uff1f<\/p>\n<p>\u5176\u5b9e\u8fd8\u5728\u521d\u671f\uff0c\u6bd5\u7adf\u521a\u5f00\u59cb\u8bad\u7ec31\u5929\u591a\u4e00\u70b9\u3002<\/p>\n<p>\u4e0d\u8fc7\u5df2\u7ecf\u80fd\u770b\u51fa\u4e24\u8005\u5728RL\u8fc7\u7a0b\u4e2d\u90fd\u6709\u4e86\u6bd4\u8f83\u660e\u663e\u7684\u80fd\u529b\u63d0\u5347\uff1a<\/p>\n<p>Pro\u7684dynsam\/avg@n\u4ece\u7b2c1\u6b65\u7684\u7ea60.565\u63d0\u5347\u81f30.614\uff0cFlash\u5219\u4ece\u7ea60.514\u63d0\u5347\u81f30.603\uff1b\u6700\u65b0\u516c\u5f00\u7684DeepSWE v1.1\u79bb\u7ebf\u8bc4\u6d4b\u4e2d\uff0cPro\u548cFlash\u5206\u522b\u8fbe\u523062.24\u548c60.77\u3002\uff08\u5bf9\u6bd4DeepSeek-Flash v1.1\u662f74.2%\uff09<\/p>\n<p>Flash\u4ece\u66f4\u4f4e\u7684\u8d77\u70b9\u5feb\u901f\u8ffd\u8fd1\uff0cPro\u76ee\u524d\u53ea\u4fdd\u6301\u5c0f\u5e45\u9886\u5148\u3002\u4e0d\u8fc7Pro\u8bad\u7ec3\u5b8c\u4e00\u4e2aStep\u66f4\u6162\uff0c\u6240\u4ee5\u6700\u65b0\u7684\u8bc4\u5206\u8fd8\u6ca1\u51fa\u6765\u3002<\/p>\n<p>\u81ea4\u6708\u4efd\u5f00\u6e90MiMo-v2.5\u540e\uff0c\u5c0f\u7c73\u6c89\u9ed8\u4e86\u8fd1\u534a\u5e74\u3002<\/p>\n<p>\u73b0\u5728\u8d1f\u8d23\u4eba\u7f57\u798f\u8389\u51fa\u6765\u89e3\u91ca\uff0c\u8fd9\u671f\u95f4\u53ea\u7814\u7a76\u4e86\u4e00\u4ef6\u4e8b\uff1a\u5f3a\u5316\u5b66\u4e60\u80fd\u6269\u5c55\u5230\u591a\u8fdc\u3002<\/p>\n<p>\u4f20\u7edf\u7684\u9884\u8bad\u7ec3scaling\u5f88\u5bb9\u6613\u7406\u89e3\uff1a\u66f4\u591a\u6570\u636e \u00d7 \u66f4\u591a\u53c2\u6570 \u00d7 \u66f4\u591a\u7b97\u529b \u2192 \u66f4\u5f3a\u6a21\u578b<\/p>\n<p>\u73b0\u5728\u524d\u6cbf\u6a21\u578b\u8d8a\u6765\u8d8a\u5173\u6ce8RL\u80fd\u4e0d\u80fd\u4e5f\u8fd9\u6837scaling\uff1f<\/p>\n<p>\u4e5f\u5c31\u8bf4\u5982\u679c\u6301\u7eed\u589e\u52a0\uff1a<\/p>\n<p>\u6bcf\u8f6e\u751f\u6210\u591a\u5c11\u8f68\u8ff9\uff1b<\/p>\n<p>\u6a21\u578b\u9762\u5bf9\u591a\u5c11\u79cd\u771f\u5b9e\u4efb\u52a1\u73af\u5883\uff1b<\/p>\n<p>\u4e3a\u5224\u65ad\u8fd9\u4e9b\u8f68\u8ff9\u597d\u4e0d\u597d\u6295\u5165\u591a\u5c11\u8ba1\u7b97\uff1b<\/p>\n<p>\u6a21\u578b\u80fd\u529b\u662f\u5426\u8fd8\u80fd\u6301\u7eed\u63d0\u9ad8\uff1f<\/p>\n<p>\u5c0f\u7c73\u7ed9\u51fa\u7684\u7b54\u6848\u662f\u6cbf\u4e09\u4e2a\u7ef4\u5ea6\u6269\u5c55\uff1a\u8bad\u7ec3\u8ba1\u7b97\u91cf\u3001\u73af\u5883\/\u6267\u884c\u6846\u67b6\u3001\u8bc4\u5206\u8ba1\u7b97\u91cf\u3002<\/p>\n<p>####\u8bad\u7ec3\u8ba1\u7b97\u91cf\u7684Scaling<\/p>\n<p>\u7b2c\u4e00\u4e2a\u7ef4\u5ea6\u6700\u76f4\u89c2\uff0c\u5c31\u662f\u628aRL\u672c\u8eab\u7684\u8ba1\u7b97\u89c4\u6a21\u505a\u5927\u3002<\/p>\n<p>MiMo-V2.6\u7cfb\u5217\u6bcf\u4e2a\u8bad\u7ec3Step\u5927\u7ea6\u4f1a\u5904\u740620\u4ebfToken\uff0c\u914d\u7f6e\u4e3a1568\u4e2aPrompt \u00d7 \u6bcf\u4e2aPrompt 16\u6761Rollout\u3002<\/p>\n<p>\u4e5f\u5c31\u662f\u8bf4\uff0c\u540c\u4e00\u9053\u9898\u5e76\u4e0d\u662f\u53ea\u8ba9\u6a21\u578b\u56de\u7b54\u4e00\u6b21\uff0c\u800c\u662f\u8ba9\u5b83\u5c1d\u8bd5\u591a\u6761\u4e0d\u540c\u7684\u89e3\u9898\u548c\u884c\u52a8\u8f68\u8ff9\uff0c\u518d\u4ece\u8fd9\u4e9b\u5c1d\u8bd5\u4e2d\u83b7\u5f97\u5f3a\u5316\u5b66\u4e60\u4fe1\u53f7\u3002<\/p>\n<p>1568\u00d716\u610f\u5473\u7740\u4e00\u8f6e\u5c31\u53ef\u80fd\u4ea7\u751f\u8d85\u8fc72.5\u4e07\u6761Rollout\u3002\u800c\u5bf9\u4e8eAgent\u4efb\u52a1\u6765\u8bf4\uff0c\u4e00\u6761Rollout\u53c8\u8fdc\u4e0d\u53ea\u662f\u201c\u4e00\u95ee\u4e00\u7b54\u201d\uff1a\u6a21\u578b\u53ef\u80fd\u8981\u7ecf\u5386\u51e0\u5341\u8f6e\u601d\u8003\u3001\u5de5\u5177\u8c03\u7528\u3001\u73af\u5883\u53cd\u9988\u548c\u518d\u6b21\u884c\u52a8\uff0c\u56e0\u6b64\u6700\u7ec8\u4e00\u4e2aStep\u7684Token\u6570\u53ef\u4ee5\u8fbe\u5230\u6570\u5341\u4ebf\u7ea7\u522b\u3002<\/p>\n<p>\u66f4\u91cd\u8981\u7684\u662f\uff0c\u6574\u4e2a\u7cfb\u7edf\u91c7\u7528\u4e86Fully Async\uff0c\u4e5f\u5c31\u662f\u5b8c\u5168\u5f02\u6b65\u7684\u6267\u884c\u65b9\u5f0f\u3002<\/p>\n<p>\u4f20\u7edf\u610f\u4e49\u4e0a\u53ef\u4ee5\u628a\u4e00\u6b21RL\u8bad\u7ec3\u60f3\u8c61\u6210\u4e00\u6761\u6574\u9f50\u7684\u6d41\u6c34\u7ebf\uff1a\u5148\u628a\u6240\u6709\u6837\u672c\u751f\u6210\u5b8c\uff0c\u518d\u7edf\u4e00\u8bc4\u5206\uff0c\u7136\u540e\u8bad\u7ec3\u6a21\u578b\uff0c\u6700\u540e\u5f00\u59cb\u4e0b\u4e00\u8f6e\u3002<\/p>\n<p>\u4f46\u5927\u89c4\u6a21Agent RL\u5f88\u96be\u8fd9\u6837\u7b49\u5f85\u3002<\/p>\n<p>\u5728MiMo\u7684\u7cfb\u7edf\u91cc\uff0c\u4e0d\u540c\u4efb\u52a1\u53ef\u4ee5\u540c\u65f6\u5904\u4e8e\u4e0d\u540c\u9636\u6bb5\uff1a\u6709\u7684Agent\u8fd8\u5728\u73af\u5883\u4e2d\u6267\u884c\u4efb\u52a1\uff0c\u6709\u7684\u5df2\u7ecf\u5b8c\u6210\u3001\u7b49\u5f85\u5224\u5206\uff0c\u6709\u7684\u6b63\u5728\u8ba1\u7b97Reward\uff0c\u8fd8\u6709\u65b0\u7684\u4efb\u52a1\u6b63\u5728\u8fdb\u5165\u7cfb\u7edf\u3002\u751f\u6210\u3001\u6267\u884c\u3001\u8bc4\u5206\u548c\u8bad\u7ec3\u4e0d\u518d\u4e25\u683c\u6392\u961f\uff0c\u800c\u662f\u7ec4\u6210\u4e86\u4e00\u5957\u6301\u7eed\u8fd0\u8f6c\u7684\u5f02\u6b65\u6d41\u6c34\u7ebf\u3002<\/p>\n<p>\u7b2c\u4e8c\u4e2a\u6269\u5c55\u65b9\u5411\u662fEnvironments and Harnesses\uff0c\u4e5f\u5c31\u662f\u8bad\u7ec3\u73af\u5883\u548c\u6267\u884c\u6846\u67b6\u7684\u6269\u5c55\u3002<\/p>\n<p>MiMo-V2.6\u505a\u7684\u662fMulti-task Agentic RL\uff1a\u4ee3\u7801\u3001\u901a\u7528\u4efb\u52a1\u3001\u89c6\u89c9\u3001Chat\u7b49\u4e0d\u540c\u7c7b\u578b\u7684\u4efb\u52a1\uff0c\u53ef\u4ee5\u88ab\u6df7\u5408\u5230\u540c\u4e00\u6b21RL Run\u4e2d\u8bad\u7ec3\uff0c\u800c\u4e14\u8fd9\u4e9b\u4efb\u52a1\u8fd8\u53ef\u4ee5\u8fd0\u884c\u5728\u4e0d\u540c\u7684Harness\u91cc\u3002<\/p>\n<p>\u4f8b\u5982\u4e00\u4e2a\u7f16\u7a0bAgent\u7684Harness\u53ef\u80fd\u5141\u8bb8\u6a21\u578b\u6253\u5f00\u7ec8\u7aef\u3001\u4fee\u6539\u4ee3\u7801\u3001\u6267\u884c\u6d4b\u8bd5\u3001\u9605\u8bfb\u62a5\u9519\uff0c\u7136\u540e\u7ee7\u7eed\u4fee\u6539\uff1b\u53e6\u4e00\u4e2a\u89c6\u89c9Agent\u9762\u5bf9\u7684\u5219\u53ef\u80fd\u662f\u4e00\u5957\u5b8c\u5168\u4e0d\u540c\u7684\u5de5\u5177\u3001\u73af\u5883\u53cd\u9988\u548c\u6210\u529f\u6761\u4ef6\u3002<\/p>\n<p>\u56e0\u6b64\uff0cMiMo\u60f3\u6269\u5c55\u7684\u9664\u4e86\u201c\u9898\u76ee\u6570\u91cf\u201d\uff0c\u8fd8\u6709\u6a21\u578b\u80fd\u591f\u8fdb\u5165\u591a\u5c11\u79cd\u73af\u5883\u3001\u4f7f\u7528\u591a\u5c11\u79cd\u5de5\u5177\u3001\u89e3\u51b3\u591a\u5c11\u79cd\u7c7b\u578b\u7684\u95ee\u9898\u3002<\/p>\n<p>\u8fd9\u4e5f\u662f\u4e3a\u4ec0\u4e48\u76f4\u64ad\u9875\u9762\u4f1a\u4e13\u95e8\u5c55\u793aBatch Composition\uff1a\u8fd9\u4e00\u680f\u5b9e\u9645\u4e0a\u662f\u5728\u544a\u8bc9\u5916\u754c\uff0c\u6bcf\u4e00\u4e2a\u8bad\u7ec3Step\u4e2d\uff0c\u6a21\u578b\u6b63\u5728\u4ece\u4ec0\u4e48\u6837\u7684\u4efb\u52a1\u548c\u73af\u5883\u91cc\u83b7\u53d6\u7ecf\u9a8c\u3002<\/p>\n<p>\u7b2c\u4e09\u4e2aScaling\u65b9\u5411\u5219\u66f4\u5bb9\u6613\u88ab\u5ffd\u7565\uff1aGrader Compute\uff0c\u4e5f\u5c31\u662f\u7ed9\u6253\u5206\u672c\u8eab\u589e\u52a0\u7b97\u529b\u3002<\/p>\n<p>\u5f3a\u5316\u5b66\u4e60\u9700\u8981\u5956\u52b1\uff0c\u4f46\u590d\u6742Agent\u4efb\u52a1\u7684\u5956\u52b1 \u5e76\u4e0d\u50cf\u6570\u5b66\u9009\u62e9\u9898\u90a3\u6837\u5bb9\u6613\u5f97\u5230\u3002<\/p>\n<p>\u4ee3\u7801\u4efb\u52a1\u8fd8\u53ef\u4ee5\u8fd0\u884cTest Case\uff1a100\u4e2a\u6d4b\u8bd5\u901a\u8fc7\u4e86\u591a\u5c11\u4e2a\uff0c\u5c31\u53ef\u4ee5\u5f62\u6210\u76f8\u5bf9\u5ba2\u89c2\u7684\u53cd\u9988\u3002<\/p>\n<p>\u4f46\u9762\u5bf9\u66f4\u5f00\u653e\u7684Agent\u4efb\u52a1\uff0c\u4ec5\u4ec5\u5224\u65ad\u201c\u6700\u7ec8\u6210\u529f\u8fd8\u662f\u5931\u8d25\u201d\u5f80\u5f80\u8fdc\u8fdc\u4e0d\u591f\uff0c\u8d1f\u8d23\u5224\u65ad\u6a21\u578b\u505a\u5f97\u597d\u4e0d\u597d\u7684\u8bc4\u5206\u8005\uff0c\u540c\u6837\u5f00\u59cb\u6d88\u8017\u8d8a\u6765\u8d8a\u591a\u8ba1\u7b97\u8d44\u6e90\u3002<\/p>\n<p>\u800c\u8fd9\u91cc\u8fd8\u6709\u4e00\u4e2a\u66f4\u52a0\u5173\u952e\u7684\u95ee\u9898Credit Assignment\uff0c\u4e5f\u5c31\u662f\u4fe1\u7528\u5206\u914d\u3002<\/p>\n<p>\u5047\u8bbe\u4e00\u4e2aAgent\u4e3a\u4e86\u5b8c\u6210\u4efb\u52a1\u8fde\u7eed\u6267\u884c\u4e8640\u6b65\uff1a\u641c\u7d22\u8d44\u6599\u3001\u6253\u5f00\u7f51\u9875\u3001\u9605\u8bfb\u4fe1\u606f\u3001\u7f16\u5199\u4ee3\u7801\u3001\u8fd0\u884c\u6d4b\u8bd5\u3001\u53d1\u73b0\u9519\u8bef\u3001\u4fee\u6539\u4ee3\u7801\uff0c\u6700\u7ec8\u6210\u529f\u5b8c\u6210\u4efb\u52a1\u3002<\/p>\n<p>\u5982\u679c\u7cfb\u7edf\u6700\u540e\u53ea\u544a\u8bc9\u6a21\u578b\u4e00\u53e5\u201c\u6210\u529f\uff0cReward=1\u201d\uff0c\u8fd9\u4e2a\u5b66\u4e60\u4fe1\u53f7\u5176\u5b9e\u975e\u5e38\u7c97\u7cd9\u3002<\/p>\n<p>\u56e0\u4e3a\u6a21\u578b\u5e76\u4e0d\u77e5\u9053\uff1a\u524d\u9762\u768440\u6b65\u4e2d\uff0c\u7a76\u7adf\u54ea\u4e9b\u52a8\u4f5c\u771f\u6b63\u5e2e\u52a9\u4e86\u6210\u529f\uff1f\u54ea\u4e9b\u6b65\u9aa4\u5176\u5b9e\u6beb\u65e0\u5fc5\u8981\uff1f\u54ea\u4e00\u6b21\u4fee\u6539\u626d\u8f6c\u4e86\u6574\u4e2a\u4efb\u52a1\uff1f\u53c8\u6709\u54ea\u4e9b\u52a8\u4f5c\u867d\u7136\u6700\u540e\u6ca1\u6709\u5bfc\u81f4\u5931\u8d25\uff0c\u5374\u5b9e\u9645\u4e0a\u662f\u4e0d\u597d\u7684\u9009\u62e9\uff1f<\/p>\n<p>MiMo\u5728\u8fd9\u6b21\u8bad\u7ec3\u4e2d\u7279\u522b\u63d0\u5230\u4e86Agentic In-group Credit Assignment\u3002\u76ee\u524d\u8fd8\u6ca1\u6709\u516c\u5e03\u5177\u4f53\u7b97\u6cd5\uff0c\u4f46\u7ed3\u5408\u201c\u540c\u4e00\u4e2aPrompt\u751f\u621016\u6761Rollout\u201d\u7684\u8bad\u7ec3\u65b9\u5f0f\uff0c\u53ef\u4ee5\u7406\u89e3\u5176\u57fa\u672c\u76ee\u6807\uff1a\u5229\u7528\u540c\u7ec4\u591a\u6761Agent\u8f68\u8ff9\u53ca\u5176\u7ed3\u679c\uff0c\u83b7\u5f97\u6bd4\u7b80\u5355\u7684\u6700\u7ec8\u6210\u8d25\u66f4\u52a0\u7ec6\u81f4\u7684\u5f3a\u5316\u5b66\u4e60\u4fe1\u53f7\u3002<\/p>\n<p>\u540c\u4e00\u7ec4\u4efb\u52a1\u5c1d\u8bd5\uff0c\u54ea\u4e9b\u884c\u4e3a\u5e94\u8be5\u83b7\u5f97\u66f4\u591a\u9f13\u52b1\u3001\u54ea\u4e9b\u5e94\u8be5\u51cf\u5c11\u9f13\u52b1\uff0c\u4ee5\u53ca\u5982\u4f55\u628a\u8fd9\u79cd\u5224\u65ad\u8f6c\u6210\u8bad\u7ec3\u4fe1\u53f7\u3002<\/p>\n<p>\u8fd9\u6837\u4e00\u6765\uff0cMiMo\u6240\u8bf4\u7684\u4e09\u4e2aScaling\u65b9\u5411\u5b9e\u9645\u4e0a\u6784\u6210\u4e86\u4e00\u5957\u5b8c\u6574\u4f53\u7cfb\uff1a<\/p>\n<p>\u6269\u5c55\u8ba1\u7b97\u91cf\uff0c\u8ba9\u6a21\u578b\u4ea7\u751f\u66f4\u591a\u7ecf\u9a8c\uff1b\u6269\u5c55\u73af\u5883\u548cHarness\uff0c\u8ba9\u6a21\u578b\u83b7\u5f97\u66f4\u52a0\u591a\u6837\u7684\u7ecf\u9a8c\uff1b\u66f4\u591a\u8bc4\u5206\u8ba1\u7b97\u91cf\uff0c\u5219\u8d1f\u8d23\u4ece\u8fd9\u4e9b\u6d77\u91cf\u7ecf\u9a8c\u4e2d\u63d0\u53d6\u66f4\u52a0\u51c6\u786e\u7684\u5b66\u4e60\u4fe1\u53f7\u3002<\/p>\n<p>\u5f53\u9884\u8bad\u7ec3\u7684Scaling\u5df2\u7ecf\u53d1\u5c55\u591a\u5e74\u4e4b\u540e\uff0c\u80fd\u5426\u628a\u6a21\u578b\u4e0e\u73af\u5883\u4ea4\u4e92\u3001\u4ea7\u751f\u7ecf\u9a8c\u3001\u8bc4\u4ef7\u7ecf\u9a8c\u518d\u5230\u5b66\u4e60\u7ecf\u9a8c\u7684\u6574\u4e2aRL\u5faa\u73af\uff0c\u4e5f\u53d8\u6210\u4e00\u53f0\u53ef\u4ee5\u6301\u7eed\u6269\u5927\u89c4\u6a21\u7684\u673a\u5668\uff1f<\/p>\n<p>\u4e0d\u8fc7\u524d\u6765\u56f4\u89c2\u7684\u5927\u4f6cViT\u5927\u4f6c\u7ed9\u51fa\u66f4\u76f4\u63a5\u7684\u7ffb\u8bd1\uff1a\u4e09\u4ef6\u4e8b\uff0c\u80cc\u540e\u90fd\u662f\u7b97\u529b\u3002<\/p>\n<p>\u53c2\u8003\u94fe\u63a5[1]https:\/\/mimo.xiaomi.com\/rl[2]https:\/\/x.com\/_LuoFuli\/status\/2100296686719610932<\/p>\n<blockquote>\n<p><strong>\u7248\u6743\u58f0\u660e\uff1a<\/strong><\/p>\n<p>\u672c\u535a\u5ba2\u90e8\u5206\u5185\u5bb9\u4e3a\u8f6c\u8f7d\u6587\u7ae0\uff0c\u65e8\u5728\u5206\u4eab\u6709\u4ef7\u503c\u7684\u4fe1\u606f\uff0c\u7248\u6743\u5f52\u539f\u4f5c\u8005\u6240\u6709\u3002<\/p>\n<p>\u8f6c\u8f7d\u4ec5\u4e3a\u4e2a\u4eba\u5b66\u4e60\u4e0e\u4ea4\u6d41\u76ee\u7684\uff0c\u4e0d\u5bf9\u6587\u7ae0\u89c2\u70b9\u8d1f\u8d23\uff0c\u4ea6\u4e0d\u7528\u4e8e\u4efb\u4f55\u5546\u4e1a\u7528\u9014\u3002<\/p>\n<p>\u5982\u6d89\u53ca\u7248\u6743\u95ee\u9898\uff0c\u8bf7\u8054\u7cfb\u672c\u4eba\u5220\u9664\u3002<\/p>\n<\/blockquote>\n","protected":false},"excerpt":{"rendered":"<p>\u5956\u52b1\u66f2\u7ebf\u3001\u663e\u5361\u6545\u969c\u5168\u516c\u5f00<\/p>\n<p>\u4ec0\u4e48\u60c5\u51b5\uff1f\u5c0f\u7c73\u65b0\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\uff0c\u76f4\u63a5\u5f00\u4e86\u73b0\u573a\u76f4\u64ad\u3002<\/p>\n<p>\u70b9\u8fdb\u53bb\u4e00\u770b\uff0c\u8fd9\u54ea\u662f\u8bad\u7ec3\u73b0\u573a\u554a\uff0c\u8fd9\u5c31\u662f\u70e7\u94b1\u73b0\u573a\uff0c\u4e00\u7728\u773c\u5c31\u662f10\u7f8e\u5200\uff0c\u4e00\u5206\u949f\u5c31\u662f4000\u591a\u5143\u4eba\u6c11\u5e01\u51fa\u53bb\u4e86\u3002<\/p>\n<p>\u5f53\u7136\u70e7\u94b1\u7684\u901f\u5ea6\u53ef\u80fd\u4e0d\u662f\u5747\u5300\u7684\uff0c\u4ecepro\u6a21\u578b\u5f00\u59cb\u8bad\u7ec3\u7b97\u8d7736\u5c0f\u65f6\uff0c\u4e24\u6b3e\u6a21\u578b\u5df2\u7ecf\u82b1\u4e86\u8d85\u8fc7108\u4e07\u7f8e\u5143\uff0c\u5e73\u5747\u6bcf\u5c0f\u65f63\u4e07\u7f8e\u5143\u3002<\/p>\n<p>\u8981\u662f\u522b\u4eba\u5e72\u8fd9\u4e8b\u5f97\u558a\u4e00\u53e5\u8bef\u95ef\u5929\u5bb6\uff0c\u4f46\u662f\u5c0f\u7c73\u5e72\u8fd9\u4e8b\u53ea\u80fd\u558a\u4e00\u53e5\u8bef\u95ef\u7c73\u5bb6\u4e86\u3002<\/p>\n<p>\u5728\u8fd9\u4e2a\u9875\u9762\u4e0a\uff0c<\/p>\n","protected":false},"author":1,"featured_media":8095,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[74],"class_list":["post-8099","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-xwzx","tag-74"],"_links":{"self":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts\/8099","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/comments?post=8099"}],"version-history":[{"count":0,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts\/8099\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/media\/8095"}],"wp:attachment":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/media?parent=8099"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/categories?post=8099"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/tags?post=8099"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}