{"id":25444,"date":"2025-04-15T10:16:29","date_gmt":"2025-04-15T02:16:29","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=25444"},"modified":"2025-04-15T10:16:29","modified_gmt":"2025-04-15T02:16:29","slug":"%e8%bf%87%e7%a8%8b%e5%a5%96%e5%8a%b1%e6%a8%a1%e5%9e%8b%e4%b9%9f%e5%8f%af%e4%bb%a5%e6%b5%8b%e8%af%95%e6%97%b6%e6%89%a9%e5%b1%95%ef%bc%9f%e6%b8%85%e5%8d%8e%e3%80%81%e4%b8%8a%e6%b5%b7ai-lab-23k%e6%95%b0","status":"publish","type":"post","link":"https:\/\/aif.amtbbs.org\/index.php\/2025\/04\/15\/25444\/","title":{"rendered":"\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u4e5f\u53ef\u4ee5\u6d4b\u8bd5\u65f6\u6269\u5c55\uff1f\u6e05\u534e\u3001\u4e0a\u6d77AI Lab 23K\u6570\u636e\u8ba91.5B\u5c0f\u6a21\u578b\u9006\u88adGPT-4o"},"content":{"rendered":"<div>\n<figure id=\"attachment_25445\" aria-describedby=\"caption-attachment-25445\" style=\"width: 300px\" class=\"wp-caption alignnone\"><img data-dominant-color=\"78aeca\" data-has-transparency=\"false\" style=\"--dominant-color: #78aeca;\" loading=\"lazy\" decoding=\"async\" class=\"not-transparent alignnone size-full wp-image-25446\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2025\/04\/425bc59670a117fc5a0532a5e8b7d2bb90e47f-300x167-1.jpg\" width=\"300\" height=\"167\" alt=\"\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2025\/04\/425bc59670a117fc5a0532a5e8b7d2bb90e47f-300x167-1.jpg 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2025\/04\/425bc59670a117fc5a0532a5e8b7d2bb90e47f-300x167-1-150x84.jpg 150w\" sizes=\"auto, (max-width: 300px) 100vw, 300px\" \/><figcaption id=\"caption-attachment-25445\" class=\"wp-caption-text\">Future medical technology controlled by AI robot using machine learning and artificial intelligence to analyze people health and give advice on health care treatment decision . 3D illustration .<\/figcaption><\/figure>\n<\/div>\n<div><\/div>\n<div class=\"article-desc\">\u6e05\u534e\u5927\u5b66\u8054\u5408\u4e0a\u6d77 AI Lab \u63d0\u51fa\u751f\u6210\u5f0f\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b \u2014\u2014GenPRM\uff0c\u5c06\u751f\u6210\u5f0f\u601d\u7ef4\u94fe\u63a8\u7406\uff08CoT\uff09\u4e0e\u4ee3\u7801\u9a8c\u8bc1\u76f8\u7ed3\u5408\uff0c\u5e76\u5f15\u5165\u6d4b\u8bd5\u65f6\u62d3\u5c55\u673a\u5236\uff0c\u4e3a\u8fc7\u7a0b\u76d1\u7763\u63a8\u7406\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\u3002<\/div>\n<div id=\"postspictures\" class=\"article-content\">\n<div id=\"container\" class=\"container am-engine\" data-v-1d7a5742=\"\" data-element=\"root\">\n<p>\u8d75\u4fed\uff0c\u5317\u4eac\u90ae\u7535\u5927\u5b66\u672c\u79d1\u4e09\u5e74\u7ea7\uff0c\u7814\u7a76\u65b9\u5411\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u3002\u5218\u6da6\u6cfd\uff0c\u6e05\u534e\u5927\u5b66\u7855\u58eb\u4e8c\u5e74\u7ea7\uff0c\u5e08\u4ece\u674e\u79c0\u6559\u6388\uff0c\u7814\u7a76\u65b9\u5411\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u4e0e\u5f3a\u5316\u5b66\u4e60\uff0c\u7279\u522b\u5173\u6ce8\u5927\u6a21\u578b\u63a8\u7406\u80fd\u529b\u589e\u5f3a\u4e0e\u6d4b\u8bd5\u65f6\u95f4\u6269\u5c55\uff0c\u5728 NeurIPS\u3001ICML\u3001ICLR\u3001AAAI \u7b49\u9876\u7ea7\u5b66\u672f\u4f1a\u8bae\u53d1\u8868\u591a\u7bc7\u8bba\u6587\uff0c\u4e2a\u4eba\u4e3b\u9875\uff1aryanliu112.github.io\u3002<\/p>\n<p>\u968f\u7740 OpenAI o1 \u548c DeepSeek R1 \u7684\u7206\u706b\uff0c\u5927\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u7684\u63a8\u7406\u80fd\u529b\u589e\u5f3a\u548c\u6d4b\u8bd5\u65f6\u6269\u5c55\uff08TTS\uff09\u53d7\u5230\u5e7f\u6cdb\u5173\u6ce8\u3002\u7136\u800c\uff0c\u5728\u590d\u6742\u63a8\u7406\u95ee\u9898\u4e2d\uff0c\u5982\u4f55\u7cbe\u51c6\u8bc4\u4f30\u6a21\u578b\u6bcf\u4e00\u6b65\u56de\u7b54\u7684\u8d28\u91cf\uff0c\u4ecd\u7136\u662f\u4e00\u4e2a\u4e9f\u5f85\u89e3\u51b3\u7684\u96be\u9898\u3002\u4f20\u7edf\u7684\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\uff08PRM\uff09\u867d\u80fd\u9a8c\u8bc1\u63a8\u7406\u6b65\u9aa4\uff0c\u4f46\u53d7\u9650\u4e8e\u6807\u91cf\u8bc4\u5206\u673a\u5236\uff0c\u96be\u4ee5\u6355\u6349\u6df1\u5c42\u903b\u8f91\u9519\u8bef\uff0c\u4e14\u5176\u5224\u522b\u5f0f\u5efa\u6a21\u65b9\u5f0f\u9650\u5236\u4e86\u6d4b\u8bd5\u65f6\u7684\u62d3\u5c55\u80fd\u529b\u3002<\/p>\n<p>\u90a3\u4e48\uff0c\u662f\u5426\u6709\u529e\u6cd5\u901a\u8fc7\u6d4b\u8bd5\u65f6\u62d3\u5c55\u63d0\u5347\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u7684\u8fc7\u7a0b\u76d1\u7763\u63a8\u7406\u80fd\u529b\u5462\uff1f<\/p>\n<p>\u4e3a\u6b64\uff0c\u6e05\u534e\u5927\u5b66\u8054\u5408\u4e0a\u6d77 AI Lab \u63d0\u51fa\u751f\u6210\u5f0f\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b \u2014\u2014GenPRM\uff0c\u5c06\u751f\u6210\u5f0f\u601d\u7ef4\u94fe\u63a8\u7406\uff08CoT\uff09\u4e0e\u4ee3\u7801\u9a8c\u8bc1\u76f8\u7ed3\u5408\uff0c\u5e76\u5f15\u5165\u6d4b\u8bd5\u65f6\u62d3\u5c55\u673a\u5236\uff0c\u4e3a\u8fc7\u7a0b\u76d1\u7763\u63a8\u7406\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\u3002\u4e0e DeepSeek \u8fd1\u671f\u53d1\u5e03\u7684\u9010\u70b9\u751f\u6210\u5956\u52b1\u6a21\u578b\uff08GRM\uff09\u7c7b\u4f3c\uff0cGenPRM \u4e5f\u901a\u8fc7\u751f\u6210\u5f0f\u5efa\u6a21\u548c\u6d4b\u8bd5\u65f6\u6269\u5c55\u589e\u5f3a\u5956\u52b1\u6a21\u578b\u7684\u63a8\u7406\u80fd\u529b\uff0c\u4f46 GenPRM \u66f4\u4e13\u6ce8\u4e8e\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\uff0c\u5f25\u8865\u4e86 GRM \u5728\u8fc7\u7a0b\u76d1\u7763\u65b9\u9762\u7684\u4e0d\u8db3\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s5.51cto.com\/oss\/202504\/14\/75acced6690dfce4b456595f5f34f6df7e1939.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<ul data-id=\"ua73dd4b-8WVMFbY0\">\n<li data-id=\"ld70c578-pEVmPEmQ\">\u8bba\u6587\u6807\u9898\uff1aGenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning<\/li>\n<li data-id=\"ld70c578-EBQsXq0O\">\u8bba\u6587\u94fe\u63a5\uff1ahttp:\/\/arxiv.org\/abs\/2504.00891<\/li>\n<li data-id=\"ld70c578-TOA92ZZd\">\u9879\u76ee\u94fe\u63a5\uff1ahttps:\/\/ryanliu112.github.io\/GenPRM<\/li>\n<li data-id=\"ld70c578-e0qtbJsv\">GitHub\uff1ahttps:\/\/github.com\/RyanLiu112\/GenPRM<\/li>\n<li data-id=\"ld70c578-OXSAfsfr\">HuggingFace\uff1ahttps:\/\/huggingface.co\/GenPRM<\/li>\n<\/ul>\n<p>\u5728 ProcessBench \u7b49\u6570\u5b66\u63a8\u7406\u57fa\u51c6\u7684\u6d4b\u8bd5\u4e2d\uff0cGenPRM \u5c55\u73b0\u51fa\u60ca\u4eba\u5b9e\u529b\uff1a\u4ec5 1.5B \u53c2\u6570\u7684\u6a21\u578b\u901a\u8fc7\u6d4b\u8bd5\u65f6\u6269\u5c55\u8d85\u8d8a GPT-4o\uff0c\u800c 7B \u53c2\u6570\u7248\u672c\u66f4\u662f\u51fb\u8d25 72B \u53c2\u6570\u7684 Qwen2.5-Math-PRM-72B\uff0c\u540c\u65f6\u8868\u73b0\u51fa\u5f3a\u5927\u7684\u6b65\u9aa4\u7ea7\u6279\u8bc4\u80fd\u529b\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s4.51cto.com\/oss\/202504\/14\/91807ee60dc94fe4d452174ab502e3fa540e29.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<h4>GenPRM\uff1a\u4ece\u8bc4\u5206\u5230\u63a8\u7406\uff0c\u518d\u5230\u6d4b\u8bd5\u65f6\u6269\u5c55<\/h4>\n<p>\u73b0\u6709\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u4f9d\u8d56\u5206\u7c7b\u5668\u5f0f\u7684\u6807\u91cf\u8bc4\u5206\uff0c\u8fd9\u79cd \u201c\u9ed1\u7bb1\u201d \u673a\u5236\u5bfc\u81f4\u4e24\u4e2a\u6838\u5fc3\u95ee\u9898\uff1a\u4e00\u662f\u65e0\u6cd5\u89e3\u91ca\u9519\u8bef\u6839\u6e90\uff0c\u4ec5\u80fd\u5224\u65ad\u6b65\u9aa4 \u201c\u5bf9\u9519\u201d\uff0c\u5374\u65e0\u6cd5\u89e3\u91ca \u201c\u4e3a\u4f55\u9519\u201d\uff0c\u4e8c\u662f\u65e0\u6cd5\u901a\u8fc7\u589e\u52a0\u6a21\u578b\u6d4b\u8bd5\u65f6\u95f4\u8ba1\u7b97\u8d44\u6e90\u63d0\u5347\u5224\u65ad\u7cbe\u5ea6\u3002<\/p>\n<p><strong>\u751f\u6210\u5f0f\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b<\/strong><\/p>\n<p>\u4e3a\u4e86\u7a81\u7834\u8fd9\u4e9b\u74f6\u9888\uff0cGenPRM \u5f15\u5165\u751f\u6210\u5f0f\u8bbe\u8ba1\uff0c\u5f7b\u5e95\u9769\u65b0\u8fc7\u7a0b\u76d1\u7763\u8303\u5f0f\uff1a<\/p>\n<ul data-id=\"ua73dd4b-OCGfG2BZ\">\n<li data-id=\"ld70c578-fmNBOTsW\">\u601d\u7ef4\u94fe\u63a8\u7406\uff1aGenPRM \u6a21\u62df\u4eba\u7c7b\u89e3\u9898\u65f6\u7684\u903b\u8f91\u63a8\u5bfc\uff0c\u5bf9\u6bcf\u4e00\u6b65\u63a8\u7406\u8fdb\u884c\u81ea\u7136\u8bed\u8a00\u5206\u6790\uff0c\u63d0\u4f9b\u900f\u660e\u3001\u53ef\u89e3\u91ca\u7684\u6b65\u9aa4\u8bc4\u4f30\u3002<\/li>\n<li data-id=\"ld70c578-Mmsb3sMf\">\u4ee3\u7801\u9a8c\u8bc1\uff1a\u4e3a\u786e\u4fdd\u63a8\u7406\u7684\u53ef\u9760\u6027\uff0cGenPRM \u8fd8\u4f1a\u751f\u6210\u5e76\u6267\u884c\u5bf9\u5e94\u6570\u5b66\u8fd0\u7b97\u7684 Python \u4ee3\u7801\uff0c\u5c06\u6587\u5b57\u63a8\u5bfc\u4e0e\u5b9e\u9645\u8ba1\u7b97\u7ed3\u679c\u4ea4\u53c9\u9a8c\u8bc1\u3002\u4f8b\u5982\uff0c\u5728\u6c42\u89e3\u4e09\u89d2\u51fd\u6570\u8868\u8fbe\u5f0f\u65f6\uff0c\u6a21\u578b\u5148\u5206\u6790\u89d2\u5ea6\u8f6c\u6362\u7684\u5408\u7406\u6027\uff0c\u518d\u901a\u8fc7\u4ee3\u7801\u8ba1\u7b97\u5177\u4f53\u6570\u503c\uff0c\u907f\u514d \u201c\u7b26\u53f7\u63a8\u5bfc\u6b63\u786e\u4f46\u8ba1\u7b97\u5931\u8bef\u201d \u7684\u60c5\u51b5\u3002<\/li>\n<\/ul>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s4.51cto.com\/oss\/202504\/14\/32e8579250242c8973402602a6e4bbc9565023.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u5176\u5956\u52b1\u63a8\u7406\u8fc7\u7a0b\u53ef\u4ee5\u8868\u793a\u4e3a\uff1a<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s9.51cto.com\/oss\/202504\/14\/68bdf0c31c6be92ad87668f15c5d5d90661f2d.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u5176\u4e2d s_t \u4e3a\u5f53\u524d\u72b6\u6001\uff0ca_t \u4e3a\u5f53\u524d\u6b65\u9aa4\uff0cv_1:t\u22121 \u548c f_1:t-1 \u5206\u522b\u4e3a\u4e4b\u524d\u6b65\u9aa4\u7684\u63a8\u7406\u8fc7\u7a0b\u548c\u4ee3\u7801\u6267\u884c\u53cd\u9988\uff0cv_t \u548c f_t \u4e3a\u5f53\u524d\u6b65\u9aa4\u7684\u63a8\u7406\u4e0e\u53cd\u9988\u3002\u8fd9\u79cd \u201c\u5148\u89e3\u91ca\u3001\u518d\u9a8c\u8bc1\u201d \u7684\u673a\u5236\u4e0d\u4ec5\u80fd\u5224\u65ad\u5bf9\u9519\uff0c\u8fd8\u80fd\u63d0\u4f9b\u6b65\u9aa4\u7ea7\u522b\u7684\u6279\u8bc4\u6539\u8fdb\u5efa\u8bae\u548c\u4e25\u8c28\u51c6\u786e\u7684\u53cd\u9988\uff0c\u5927\u5e45\u63d0\u5347\u4e86\u8fc7\u7a0b\u76d1\u7763\u7684\u6df1\u5ea6\u548c\u5b9e\u7528\u6027\u3002<\/p>\n<p><strong>\u6d4b\u8bd5\u65f6\u6269\u5c55<\/strong><\/p>\n<p>\u5728\u63a8\u7406\u9636\u6bb5\uff0cGenPRM \u901a\u8fc7\u5e76\u884c\u91c7\u6837 N \u6761\u63a8\u7406\u8def\u5f84\uff0c\u7efc\u5408\u591a\u6761\u8def\u5f84\u7684\u5956\u52b1\u503c\u5e76\u53d6\u5e73\u5747\uff0c\u5f97\u5230\u6700\u7ec8\u5956\u52b1\uff1a<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s6.51cto.com\/oss\/202504\/14\/07a4fff3845493fce4e3785641edbe080f33fc.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u8fd9\u79cd\u7b56\u7565\u5145\u5206\u5229\u7528\u989d\u5916\u8ba1\u7b97\u8d44\u6e90\uff0c\u8fdb\u4e00\u6b65\u63d0\u5347\u8bc4\u4f30\u7cbe\u5ea6\uff0c\u4f7f\u5c0f\u6a21\u578b\u4e5f\u80fd\u5728\u590d\u6742\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\u3002<\/p>\n<h4>\u6570\u636e\u9ad8\u6548\uff1a23K \u6837\u672c\u80cc\u540e\u7684\u5408\u6210\u79d8\u5bc6<\/h4>\n<p>GenPRM \u7684\u53e6\u4e00\u4e2a\u4eae\u70b9\u662f\u4ec5\u4f7f\u7528 23K \u8bad\u7ec3\u6837\u672c\u5c31\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u6027\u80fd\uff0c\u8fdc\u5c11\u4e8e\u8bb8\u591a\u6a21\u578b\u52a8\u8f84\u6570\u5341\u4e07\u7ea7\u7684\u6570\u636e\u91cf\uff08\u5982 PRM800K \u9700 80 \u4e07\u4eba\u5de5\u6807\u6ce8\uff09\uff0c\u5176\u9ad8\u6548\u6027\u6e90\u4e8e\u72ec\u7279\u7684\u6570\u636e\u5408\u6210\u65b9\u6cd5\uff0c\u7ed3\u5408\u76f8\u5bf9\u8fdb\u6b65\u4f30\u8ba1\uff08RPE\uff09\u548c\u4ee3\u7801\u9a8c\u8bc1\uff0c\u751f\u6210\u9ad8\u8d28\u91cf\u7684\u8fc7\u7a0b\u76d1\u7763\u6570\u636e\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s5.51cto.com\/oss\/202504\/14\/c78d3ef88977447b77a762dceefecc9a8411e0.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p><strong>\u901a\u8fc7\u76f8\u5bf9\u8fdb\u6b65\u4f30\u8ba1\u6539\u8fdb\u786c\u4f30\u8ba1<\/strong><\/p>\n<p>\u4f20\u7edf\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u901a\u8fc7\u8499\u7279\u5361\u7f57\uff08MC\uff09\u5206\u6570\u8fdb\u884c\u786c\u4f30\u8ba1\uff0c\u7814\u7a76\u8005\u89c2\u5bdf\u5230\u5c3d\u7ba1\u8bb8\u591a\u6b65\u9aa4\u7684 MC \u5206\u6570\u5927\u4e8e 0\uff0c\u4f46\u8fd9\u4e9b\u6b65\u9aa4\u662f\u5374\u5b58\u5728\u9519\u8bef\u3002RPE \u901a\u8fc7\u6bd4\u8f83\u5f53\u524d\u72b6\u6001\u548c\u4e0a\u4e00\u72b6\u6001\u7684 MC \u5206\u6570\uff0c\u7528 \u201c\u8fdb\u6b65\u5e45\u5ea6\u201d \u8bc4\u4f30\u6bcf\u6b65\u8d28\u91cf\uff0c\u6bd4\u4f20\u7edf\u786c\u6807\u7b7e\u66f4\u51c6\u786e\u3002\u5176\u5f62\u5f0f\u5316\u5982\u4e0b\uff1a<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s2.51cto.com\/oss\/202504\/14\/713302756cf5925df8b121ec56dd52333c6d14.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u5176\u4e2d\uff0cMC (s_t, a_t) \u8868\u793a\u5f53\u524d\u6b65\u9aa4\u7684\u8499\u7279\u5361\u7f57\u5206\u6570\uff0cMC (s_t) \u8868\u793a\u4e0a\u4e00\u6b65\u9aa4\u7684\u8499\u7279\u5361\u7f57\u5206\u6570\u3002\u82e5\u8fdb\u6b65\u5e45\u5ea6\u4f4e\u4e8e\u9608\u503c\uff08\u03f5=0.8\uff09\uff0c\u5219\u5224\u5b9a\u6b65\u9aa4\u65e0\u6548\uff1b\u82e5\u9996\u6b65\u9519\u8bef\uff08MC \u4e3a 0\uff09\uff0c\u540e\u7eed\u6b65\u9aa4\u5206\u6570\u5f52\u96f6\u3002\u8fd9\u79cd\u65b9\u6cd5\u663e\u8457\u63d0\u5347\u6807\u7b7e\u51c6\u786e\u6027\uff0c\u907f\u514d\u4e86\u786c\u4f30\u8ba1\u7684\u8bef\u5224\u3002<\/p>\n<p><strong>\u4ee3\u7801\u9a8c\u8bc1\u9a71\u52a8\u7684\u6570\u636e\u5408\u6210<\/strong><\/p>\n<p>\u7814\u7a76\u8005\u5229\u7528 QwQ-32B \u6a21\u578b\u5408\u6210 CoT \u548c\u4ee3\u7801\u9a8c\u8bc1\u63a8\u7406\u6570\u636e\uff0c\u901a\u8fc7\u5728 Python \u73af\u5883\u4e2d\u771f\u5b9e\u6267\u884c\u4ee3\u7801\u91cd\u590d\u68c0\u9a8c CoT \u63a8\u7406\u8fc7\u7a0b\u3002\u4f7f\u7528\u5171\u8bc6\u8fc7\u6ee4\uff08\u8fc7\u6ee4\u7387 51%\uff09\uff0c\u4fdd\u7559\u9ad8\u8d28\u91cf\u8fc7\u7a0b\u76d1\u7763\u6570\u636e\uff0c\u6700\u7ec8\u5f97\u5230 23K \u8bad\u7ec3\u6570\u636e\u96c6\u3002<\/p>\n<h4>\u6d4b\u8bd5\u65f6\u6269\u5c55\uff1a\u5c0f\u6a21\u578b\u7684\u9006\u88ad<\/h4>\n<p>\u5728 ProcessBench \u8fc7\u7a0b\u76d1\u7763\u57fa\u51c6\u6d4b\u8bd5\u4e2d\uff0cGenPRM \u5c55\u73b0\u51fa\u663e\u8457\u4f18\u52bf\uff1a<\/p>\n<ul data-id=\"ua73dd4b-TYLcFiLL\">\n<li data-id=\"ld70c578-uu0nWorB\">\u4ec5\u7528 23K \u8bad\u7ec3\u6570\u636e\u7684 1.5B GenPRM\uff0c\u901a\u8fc7\u591a\u6570\u6295\u7968\uff08Maj@8\uff09\u7684\u6d4b\u8bd5\u65f6\u8ba1\u7b97\u6269\u5c55\u7b56\u7565\uff0c\u5176 F1 \u5206\u6570\u8d85\u8d8a GPT-4o\uff1b<\/li>\n<li data-id=\"ld70c578-OM2ZkW2D\">7B \u7248\u672c\u7684 GenPRM \u4ee5 80.5% \u7684 F1 \u5206\u6570\u4e00\u4e3e\u8d85\u8fc7 72B \u53c2\u6570\u7684 Qwen2.5-Math-PRM-72B\u3002<\/li>\n<\/ul>\n<p>\u8fd9\u4e00\u7ed3\u679c\u8bc1\u660e\uff0c\u6d4b\u8bd5\u65f6\u6269\u5c55\u80fd\u6709\u6548\u653e\u5927\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u7684\u80fd\u529b\uff0c\u4f7f\u5c0f\u6a21\u578b\u5b9e\u73b0\u6027\u80fd\u98de\u8dc3\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s4.51cto.com\/oss\/202504\/14\/847d687616acd1923eb2904c03b426c3ec0985.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u6b64\u5916\uff0cGenPRM \u540c\u6837\u9002\u7528\u4e8e\u7b56\u7565\u6a21\u578b\u6d4b\u8bd5\u65f6\u6269\u5c55\u3002\u901a\u8fc7 Best-of-N \u5b9e\u9a8c\uff0cGenPRM-7B \u5c55\u73b0\u51fa\u76f8\u6bd4\u4e8e\u57fa\u7ebf\u65b9\u6cd5\u66f4\u52a0\u4f18\u5f02\u7684\u7b5b\u9009\u80fd\u529b\uff0c\u5e76\u53ef\u901a\u8fc7\u6d4b\u8bd5\u65f6\u6269\u5c55\u8fdb\u4e00\u6b65\u589e\u5f3a\u8fc7\u7a0b\u76d1\u7763\u80fd\u529b\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s5.51cto.com\/oss\/202504\/14\/d88815d67776f5a8e4f077c28f79bfc5cf8694.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<h4>\u4ece\u9a8c\u8bc1\u5668\u5230\u6279\u8bc4\u8005\uff1a\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u65b0\u8303\u5f0f<\/h4>\n<p>GenPRM \u4e0d\u4ec5\u80fd\u5f53 \u201c\u88c1\u5224\u201d\uff0c\u4f5c\u4e3a\u9a8c\u8bc1\u5668\uff08Verifier\uff09\u7b5b\u9009\u7b54\u6848\uff0c\u8fd8\u80fd\u5f53 \u201c\u6559\u7ec3\u201d\uff0c\u4f5c\u4e3a\u6b65\u9aa4\u7ea7\u522b\u7684\u6279\u8bc4\u6a21\u578b\uff08Critic\uff09\u6307\u5bfc\u7b56\u7565\u6a21\u578b\u8fed\u4ee3\u4f18\u5316\u539f\u59cb\u56de\u7b54\u3002\u5b9e\u9a8c\u8868\u660e\uff0cGenPRM \u901a\u8fc7 3 \u8f6e\u53cd\u9988\u5c06\u7b56\u7565\u6a21\u578b\u7684\u56de\u7b54\u51c6\u786e\u7387\u4ece 45.7% \u63d0\u5347\u81f3 51.5%\uff0c\u6027\u80fd\u63d0\u5347\u8fbe\u5230\u57fa\u7ebf\u65b9\u6cd5\u7684 3.4 \u500d\u3002<\/p>\n<p><img decoding=\"async\" title=\"\u56fe\u7247\" src=\"https:\/\/s2.51cto.com\/oss\/202504\/14\/57195d507d235642c263332bc48bd8f77bb0ed.webp\" alt=\"\u56fe\u7247\" data-type=\"inline\" \/><\/p>\n<p>\u8fd9\u79cd \u201c\u751f\u6210 &#8211; \u6279\u8bc4 &#8211; \u53cd\u601d\u201d \u7684\u95ed\u73af\uff0c\u9a8c\u8bc1\u4e86 GenPRM \u4e0d\u4ec5\u53ef\u4ee5\u4f5c\u4e3a\u9a8c\u8bc1\u5668\u9a8c\u8bc1\u7b54\u6848\u7684\u51c6\u786e\u6027\uff0c\u8fd8\u53ef\u4ee5\u4f5c\u4e3a\u6279\u8bc4\u8005\uff0c\u4e3a\u6a21\u578b\u5b8c\u5584\u81ea\u8eab\u8f93\u51fa\u63d0\u4f9b\u9010\u6b65\u5173\u952e\u6307\u5bfc\uff0c\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u7684\u81ea\u6211\u6539\u8fdb\u63d0\u4f9b\u4e86\u53ef\u89e3\u91ca\u7684\u6280\u672f\u8def\u5f84\u3002<\/p>\n<p>\u7814\u7a76\u8005\u5df2\u5f00\u6e90\u4ee3\u7801\u3001\u6a21\u578b\u53ca 23K \u8bad\u7ec3\u6570\u636e\u96c6\u3002\u8be5\u5de5\u4f5c\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u7684\u53ef\u89e3\u91ca\u8fc7\u7a0b\u76d1\u7763\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\uff0c\u672a\u6765\u53ef\u6269\u5c55\u81f3\u4ee3\u7801\u751f\u6210\u3001\u591a\u6a21\u6001\u63a8\u7406\u7b49\u9886\u57df\u3002<\/p>\n<\/div>\n<p>\u6587\u7ae0\u6765\u81ea\uff1a51CTO<\/p>\n<\/div>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_25444\" class=\"pvc_stats total_only  \" data-element-id=\"25444\" 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:\/\/aif.amtbbs.org\/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>\u6e05\u534e\u5927\u5b66\u8054\u5408\u4e0a\u6d77 AI Lab \u63d0\u51fa\u751f\u6210\u5f0f\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b \u2014\u2014GenPRM\uff0c\u5c06\u751f\u6210\u5f0f\u601d\u7ef4\u94fe\u63a8\u7406\uff08CoT\uff09\u4e0e\u4ee3\u7801\u9a8c [&hellip;]<\/p>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_25444\" class=\"pvc_stats total_only  \" data-element-id=\"25444\" 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:\/\/aif.amtbbs.org\/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":25446,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[23,17,20,80],"tags":[124,887,820],"class_list":["post-25444","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-23","category-17","category-20","category-80","tag-ai","tag-887","tag-820"],"_links":{"self":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25444","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/comments?post=25444"}],"version-history":[{"count":1,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25444\/revisions"}],"predecessor-version":[{"id":25447,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25444\/revisions\/25447"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media\/25446"}],"wp:attachment":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media?parent=25444"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/categories?post=25444"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/tags?post=25444"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}