{"id":29457,"date":"2025-02-23T14:36:11","date_gmt":"2025-02-23T06:36:11","guid":{"rendered":"https:\/\/www.1ai.net\/?p=29457"},"modified":"2025-02-23T14:36:11","modified_gmt":"2025-02-23T06:36:11","slug":"openai-%e5%91%98%e5%b7%a5%e5%85%ac%e5%bc%80%e6%8c%87%e8%b4%a3-xai%ef%bc%9agrok-3-%e5%9f%ba%e5%87%86%e6%b5%8b%e8%af%95%e7%bb%93%e6%9e%9c%e5%85%b7%e6%9c%89%e8%af%af%e5%af%bc%e6%80%a7","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/29457.html","title":{"rendered":"OpenAI Employees Publicly Accuse xAI: Grok 3 Benchmark Results Are Misleading"},"content":{"rendered":"<p>2 \u6708 23 \u65e5\u6d88\u606f\uff0c\u672c\u5468\uff0c<a href=\"https:\/\/www.1ai.net\/en\/tag\/openai\" title=\"\u3010\u67e5\u770b\u542b\u6709[OpenAI]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">OpenAI<\/a> \u7684\u4e00\u540d\u5458\u5de5\u516c\u5f00\u6307\u8d23\u57c3\u9686\u30fb\u9a6c\u65af\u514b\u65d7\u4e0b\u7684 <a href=\"https:\/\/www.1ai.net\/en\/tag\/xai\" title=\"\u3010\u67e5\u770b\u542b\u6709[xAI]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">xAI<\/a> \u516c\u53f8\uff0c\u79f0\u5176\u53d1\u5e03\u7684\u6700\u65b0 AI \u6a21\u578b Grok 3 \u7684\u57fa\u51c6\u6d4b\u8bd5\u7ed3\u679c\u5177\u6709\u8bef\u5bfc\u6027\u3002\u5bf9\u6b64\uff0cxAI \u7684\u8054\u5408\u521b\u59cb\u4eba\u4f0a\u6208\u5c14\u30fb\u5df4\u5e03\u4ec0\u91d1\uff08Igor Babushkin\uff09\u5219\u575a\u79f0\u516c\u53f8\u5e76\u65e0\u4e0d\u5f53\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-29458\" title=\"7ba06a91j00ss4ib9001cd000ez00ewp\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/02\/7ba06a91j00ss4ib9001cd000ez00ewp.jpg\" alt=\"7ba06a91j00ss4ib9001cd000ez00ewp\" width=\"539\" height=\"536\" \/><\/p>\n<p>xAI \u5728\u5176\u535a\u5ba2\u4e0a\u53d1\u5e03\u4e86\u4e00\u5f20\u56fe\u8868\uff0c\u5c55\u793a\u4e86 Grok 3 \u5728 AIME 2025\uff08\u4e00\u9879\u8fd1\u671f\u9080\u8bf7\u5236\u6570\u5b66\u8003\u8bd5\u4e2d\u7684\u9ad8\u96be\u5ea6\u6570\u5b66\u9898\u96c6\uff09\u4e0a\u7684\u8868\u73b0\u3002\u5c3d\u7ba1\u4e00\u4e9b\u4e13\u5bb6\u8d28\u7591 AIME \u4f5c\u4e3a AI \u57fa\u51c6\u7684\u6709\u6548\u6027\uff0c\u4f46 AIME 2025 \u53ca\u5176\u65e9\u671f\u7248\u672c\u4ecd\u88ab\u5e7f\u6cdb\u7528\u4e8e\u8bc4\u4f30\u6a21\u578b\u7684\u6570\u5b66\u80fd\u529b\u3002<\/p>\n<p>1AI\u6ce8\u610f\u5230\uff0cxAI \u7684\u56fe\u8868\u663e\u793a\uff0cGrok 3 \u7684\u4e24\u4e2a\u7248\u672c \u2014\u2014Grok 3 Reasoning Beta \u548c Grok 3 mini Reasoning\u2014\u2014 \u5728 AIME 2025 \u4e0a\u7684\u8868\u73b0\u8d85\u8fc7\u4e86 OpenAI \u5f53\u524d\u6700\u5f3a\u7684\u53ef\u7528\u6a21\u578b o3-mini-high\u3002\u7136\u800c\uff0cOpenAI \u7684\u5458\u5de5\u5f88\u5feb\u5728 X \u5e73\u53f0\u4e0a\u6307\u51fa\uff0c<strong>xAI \u7684\u56fe\u8868\u5e76\u672a\u5305\u542b o3-mini-high \u5728\u201ccons@64\u201d\u6761\u4ef6\u4e0b\u7684 AIME 2025 \u5f97\u5206<\/strong>\u3002<\/p>\n<p>\u201ccons@64\u201d\u662f\u6307\u201cconsensus@64\u201d\uff0c\u5373\u5141\u8bb8\u6a21\u578b\u5728\u57fa\u51c6\u6d4b\u8bd5\u4e2d\u5bf9\u6bcf\u4e2a\u95ee\u9898\u5c1d\u8bd5 64 \u6b21\uff0c\u5e76\u5c06\u51fa\u73b0\u9891\u7387\u6700\u9ad8\u7684\u7b54\u6848\u4f5c\u4e3a\u6700\u7ec8\u7b54\u6848\u3002\u53ef\u60f3\u800c\u77e5\uff0c\u8fd9\u79cd\u65b9\u5f0f\u5f80\u5f80\u4f1a\u663e\u8457\u63d0\u5347\u6a21\u578b\u7684\u57fa\u51c6\u6d4b\u8bd5\u5206\u6570\uff0c\u5982\u679c\u56fe\u8868\u4e2d\u7701\u7565\u8fd9\u4e00\u6570\u636e\uff0c\u5c31\u53ef\u80fd\u8ba9\u4eba\u8bef\u4ee5\u4e3a\u67d0\u4e2a\u6a21\u578b\u7684\u8868\u73b0\u4f18\u4e8e\u53e6\u4e00\u6a21\u578b\uff0c\u800c\u5b9e\u9645\u60c5\u51b5\u672a\u5fc5\u5982\u6b64\u3002<\/p>\n<p>\u5728 AIME 2025 \u7684\u201c@1\u201d\u6761\u4ef6\u4e0b\uff08\u5373\u6a21\u578b\u9996\u6b21\u5c1d\u8bd5\u7684\u5f97\u5206\uff09\uff0cGrok 3 Reasoning Beta \u548c Grok 3 mini Reasoning \u7684\u5f97\u5206\u4f4e\u4e8e o3-mini-high\u3002Grok 3 Reasoning Beta \u7684\u8868\u73b0\u4e5f\u7565\u4f4e\u4e8e OpenAI \u7684 o1 \u6a21\u578b\u5728\u201c\u4e2d\u7b49\u8ba1\u7b97\u201d\u8bbe\u7f6e\u4e0b\u7684\u5f97\u5206\u3002<strong>\u7136\u800c\uff0cxAI \u4ecd\u5728\u5ba3\u4f20 Grok 3 \u4e3a\u201c\u4e16\u754c\u4e0a\u6700\u806a\u660e\u7684 AI\u201d\u3002<\/strong><\/p>\n<p>\u5df4\u5e03\u4ec0\u91d1\u5728 X \u5e73\u53f0\u4e0a\u8fa9\u79f0\uff0c<strong>OpenAI \u8fc7\u53bb\u4e5f\u66fe\u53d1\u5e03\u8fc7\u7c7b\u4f3c\u7684\u8bef\u5bfc\u6027\u57fa\u51c6\u6d4b\u8bd5\u56fe\u8868\u3002\u5c3d\u7ba1\u8fd9\u4e9b\u56fe\u8868\u662f\u7528\u4e8e\u6bd4\u8f83\u5176\u81ea\u8eab\u6a21\u578b\u7684\u8868\u73b0\u3002<\/strong><\/p>\n<p>\u5728\u8fd9\u573a\u4e89\u8bae\u4e2d\uff0c\u4e00\u4f4d\u4e2d\u7acb\u7684\u7b2c\u4e09\u65b9\u91cd\u65b0\u7ed8\u5236\u4e86\u4e00\u5f20\u66f4\u4e3a\u201c\u51c6\u786e\u201d\u7684\u56fe\u8868\uff1a<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-29459\" title=\"8caa4b75j00ss4i80007cd000v900qxp\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/02\/8caa4b75j00ss4i80007cd000v900qxp.jpg\" alt=\"8caa4b75j00ss4i80007cd000v900qxp\" width=\"1125\" height=\"969\" \/><\/p>\n<p>\u4f46\u6b63\u5982 AI \u7814\u7a76\u5458\u5185\u68ee\u30fb\u5170\u4f2f\u7279\uff08Nathan Lambert\uff09\u5728\u4e00\u7bc7\u6587\u7ae0\u4e2d\u6307\u51fa\u7684\uff0c\u6216\u8bb8\u6700\u91cd\u8981\u7684\u6307\u6807\u4ecd\u7136\u672a\u77e5\uff1a\u6bcf\u4e2a\u6a21\u578b\u8fbe\u5230\u6700\u4f73\u5206\u6570\u6240\u9700\u7684\u8ba1\u7b97\uff08\u548c\u91d1\u94b1\uff09\u6210\u672c\u3002\u8fd9\u6070\u6070\u8868\u660e\uff0c\u5927\u591a\u6570 AI \u57fa\u51c6\u6d4b\u8bd5\u5728\u4f20\u8fbe\u6a21\u578b\u7684\u5c40\u9650\u6027\u548c\u4f18\u52bf\u65b9\u9762\u4ecd\u7136\u5b58\u5728\u5f88\u5927\u7684\u4e0d\u8db3\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Feb. 23 (Bloomberg) -- This week, an OpenAI employee publicly accused Elon Musk's xAI of releasing misleading benchmark results for its latest AI model, Grok 3. In response, xAI co-founder Igor Babushkin insisted the company did nothing wrong. On its blog, xAI posted a chart showing Grok 3's performance on AIME 2025, a collection of difficult math problems from a recent invitational math exam. While some experts have questioned the validity of AIME as an AI benchmark, AIME 2025 and its earlier versions are still widely used to assess the mathematical capabilities of models. 1AI<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[3311,190,356],"collection":[],"class_list":["post-29457","post","type-post","status-publish","format-standard","hentry","category-news","tag-grok-3","tag-openai","tag-xai"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/29457","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=29457"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/29457\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=29457"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=29457"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=29457"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=29457"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}