{"id":20909,"date":"2026-11-07T17:21:22","date_gmt":"2026-11-07T09:21:22","guid":{"rendered":"https:\/\/dguan.com.cn\/ai\/?p=20909"},"modified":"2026-11-07T17:21:22","modified_gmt":"2026-11-07T09:21:22","slug":"%e6%8a%a2%e7%a5%a8%e5%92%8c%e4%bf%84%e7%bd%97%e6%96%af%e6%96%b9%e5%9d%97%e9%9a%be%e5%80%92%e6%97%97%e8%88%b0%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%9arealtime-bench-%e6%8c%82%e9%92%9f%e6%b5%8b%e8%af%95","status":"publish","type":"post","link":"https:\/\/dguan.com.cn\/ai\/%e6%8a%a2%e7%a5%a8%e5%92%8c%e4%bf%84%e7%bd%97%e6%96%af%e6%96%b9%e5%9d%97%e9%9a%be%e5%80%92%e6%97%97%e8%88%b0%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%9arealtime-bench-%e6%8c%82%e9%92%9f%e6%b5%8b%e8%af%95\/","title":{"rendered":"\u62a2\u7968\u548c\u4fc4\u7f57\u65af\u65b9\u5757\u96be\u5012\u65d7\u8230\u5927\u6a21\u578b\uff1aRealtime-Bench \u6302\u949f\u6d4b\u8bd5\u6700\u9ad8\u4ec5 44%"},"content":{"rendered":"<p>\u82f1\u4f1f\u8fbe\u8054\u5408\u5f97\u514b\u8428\u65af\u5927\u5b66\u5965\u65af\u6c40\u5206\u6821\u7b49\u673a\u6784\u7684\u7814\u7a76\u56e2\u961f\uff0c\u5728 arXiv \u53d1\u5e03\u4e86\u4e00\u9879\u540d\u4e3a Realtime-Bench \u7684\u5b9e\u65f6\u4ea4\u4e92\u57fa\u51c6\uff0c\u76f4\u6307\u5f53\u524d\u5927\u6a21\u578b\u8bc4\u6d4b\u4f53\u7cfb\u7684\u4e00\u4e2a\u76f2\u533a\uff1a\u73af\u5883\u4e0d\u4f1a\u505c\u4e0b\u6765\u7b49\u6a21\u578b\u601d\u8003\u3002\u8fd9\u9879\u5de5\u4f5c\u7531 UT-Austin\u3001\u4fc4\u4ea5\u4fc4\u5dde\u7acb\u5927\u5b66\u3001UIUC\u3001\u897f\u5317\u5927\u5b66\u4e0e\u82f1\u4f1f\u8fbe\u3001Steadyworks \u5171\u540c\u5b8c\u6210\uff0c\u8bba\u6587\u4e8e 10 \u6708 15 \u65e5\u63d0\u4ea4\u300110 \u6708 22 \u65e5\u66f4\u65b0\u81f3\u7b2c\u4e09\u7248\uff0c11 \u6708\u521d\u5728\u4e2d\u6587\u6280\u672f\u793e\u533a\u5f15\u53d1\u5e7f\u6cdb\u8ba8\u8bba\u3002<\/p>\n<p>\u57fa\u51c6\u7684\u8bbe\u8ba1\u601d\u8def\u5f88\u76f4\u63a5\u3002Realtime-Bench \u51c6\u5907\u4e86 50 \u4e2a\u52a8\u6001\u4efb\u52a1\u2014\u201430 \u4e2a\u7f51\u9875\u4efb\u52a1\u8986\u76d6\u51fa\u884c\u3001\u7535\u5546\u3001\u670d\u52a1\u3001\u5b89\u5168\u56db\u7c7b\u573a\u666f\uff0c20 \u4e2a\u6e38\u620f\u4efb\u52a1\u8986\u76d6\u8857\u673a\u3001\u89e3\u8c1c\u3001\u7b56\u7565\u3001\u6a21\u62df\u56db\u7c7b\u2014\u2014\u5171\u540c\u7279\u70b9\u662f\u73af\u5883\u968f\u65f6\u95f4\u81ea\u884c\u63a8\u8fdb\uff0c\u884c\u52a8\u673a\u4f1a\u7a0d\u7eb5\u5373\u901d\u3002\u6a21\u578b\u9700\u8981\u540c\u65f6\u5177\u5907\u4e09\u79cd\u80fd\u529b\uff1a\u5224\u65ad\u4fe1\u606f\u662f\u5426\u8fc7\u65f6\u7684\u65f6\u95f4\u611f\u77e5\uff0c\u6839\u636e\u5269\u4f59\u65f6\u95f4\u51b3\u5b9a\u601d\u8003\u591a\u4e45\u7684\u81ea\u9002\u5e94\u63a8\u7406\uff0c\u4ee5\u53ca\u628a\u63e1\u7b49\u5f85\u8fd8\u662f\u7acb\u5373\u51fa\u624b\u7684\u884c\u52a8\u65f6\u673a\u3002\u4ee5\u62a2\u7968\u4e3a\u4f8b\uff0c\u6a21\u578b\u8981\u5148\u53d1\u73b0\u653e\u7968\uff0c\u518d\u6309\u8d2d\u4e70\u7a97\u53e3\u7684\u5269\u4f59\u65f6\u957f\u63a7\u5236\u63a8\u7406\u957f\u5ea6\uff0c\u4efb\u4f55\u4e00\u4e2a\u73af\u8282\u8fdf\u4e86\u90fd\u4f1a\u9519\u8fc7\u673a\u4f1a\u3002<\/p>\n<p>\u4e3a\u4e86\u628a\u300c\u670d\u52a1\u5668\u6162\u300d\u548c\u300c\u6a21\u578b\u4e0d\u4f1a\u62a2\u65f6\u95f4\u300d\u533a\u5206\u5f00\uff0c\u7814\u7a76\u56e2\u961f\u8bbe\u8ba1\u4e86\u4e09\u79cd\u65f6\u949f\u534f\u8bae\uff1aWall Clock \u6309\u771f\u5b9e\u63a8\u7406\u5ef6\u8fdf\u63a8\u8fdb\u73af\u5883\uff0cPaused Clock \u5728\u6a21\u578b\u601d\u8003\u671f\u95f4\u51bb\u7ed3\u4e16\u754c\u4f5c\u4e3a\u7406\u60f3\u5316\u5bf9\u7167\uff0cToken Clock \u5219\u6309\u751f\u6210\u7684 token \u6570\u63a8\u8fdb\u6a21\u62df\u65f6\u95f4\uff0c\u5265\u79bb API \u670d\u52a1\u901f\u5ea6\u7684\u5f71\u54cd\u3002\u8bc4\u6d4b\u6307\u6807\u91c7\u7528 Pass^3\uff0c\u5373\u540c\u4e00\u4efb\u52a1\u4e09\u6b21\u72ec\u7acb\u8fd0\u884c\u5168\u90e8\u6210\u529f\u7684\u6982\u7387\u2014\u2014\u56e0\u4e3a\u5b9e\u65f6\u4ea4\u4e92\u4e00\u65e6\u5931\u8d25\u65e0\u6cd5\u91cd\u6765\u3002<\/p>\n<p>\u7ed3\u679c\u7ed9\u5f53\u4e0b\u6700\u70ed\u8877\u7684\u300c\u52a0\u5927\u63a8\u7406\u529b\u5ea6\u300d\u8def\u7ebf\u6cfc\u4e86\u4e00\u76c6\u51b7\u6c34\u3002\u5728\u6309\u771f\u5b9e\u5ef6\u8fdf\u63a8\u8fdb\u7684\u6302\u949f\u6761\u4ef6\u4e0b\uff0c\u8868\u73b0\u6700\u597d\u7684 GPT-6-Astra \u5355\u6b21\u6210\u529f\u7387\u4e5f\u53ea\u6709 44%\uff0c\u8fde\u7eed\u4e09\u6b21\u5168\u8fc7\u7684 Pass^3 \u66f4\u662f\u8dcc\u5230 36%\u3002\u66f4\u8010\u4eba\u5bfb\u5473\u7684\u662f\u63a8\u7406\u5f3a\u5ea6\u7684\u53cc\u5411\u6548\u5e94\uff1a\u65f6\u95f4\u51bb\u7ed3\u65f6\uff0c\u63d0\u9ad8\u63a8\u7406\u5f3a\u5ea6\u8ba9\u516d\u4e2a\u53d7\u6d4b\u6a21\u578b\u5168\u90e8\u6da8\u5206\uff0cGPT-5.6 Sol \u4ece 42% \u5347\u5230 58%\uff1b\u53ef\u4e00\u65e6\u65f6\u95f4\u7167\u5e38\u6d41\u901d\uff0c\u540c\u6837\u7684\u64cd\u4f5c\u5374\u8ba9 Claude Opus 5 \u5728 Token Clock \u4e0b\u4ece 44% \u8dcc\u5230 30%\uff0c\u800c Kimi-K3 \u53cd\u800c\u4ece 24% \u5347\u5230 36%\u3002\u4f4e\u5ef6\u8fdf\u4e5f\u4e0d\u4fdd\u9669\u2014\u2014Jev 1.13 \u662f\u552f\u4e00\u538b\u8fdb 1 \u79d2\u7684\u6a21\u578b\uff0cPass^3 \u5374\u53ea\u6709 8%\uff0c\u8bf4\u660e\u5feb\u800c\u4e0d\u6df1\u540c\u6837\u4e0d\u884c\u3002<\/p>\n<p>\u8fd9\u7ec4\u6570\u636e\u5b9e\u9645\u4e0a\u662f\u5bf9 test-time scaling \u8def\u7ebf\u7684\u4e00\u6b21\u538b\u529b\u6d4b\u8bd5\u3002\u8fc7\u53bb\u4e24\u5e74\uff0c\u884c\u4e1a\u9ed8\u8ba4\u300c\u56de\u5408\u5236\u300d\u8bc4\u6d4b\uff1a\u73af\u5883\u9759\u6b62\u7b49\u5f85\u6a21\u578b\u5410\u5b8c\u51e0\u5343\u4e2a\u63a8\u7406 token \u518d\u6267\u884c\u4e0b\u4e00\u6b65\uff0c\u62c9\u957f\u601d\u7ef4\u94fe\u51e0\u4e4e\u603b\u80fd\u6362\u5230\u66f4\u9ad8\u7684\u5206\u6570\u3002\u800c Realtime-Bench \u7684\u5bf9\u6bd4\u663e\u793a\uff0c\u65f6\u95f4\u6682\u505c\u65f6\u9876\u5c16\u6a21\u578b\u7684\u6210\u529f\u7387\u80fd\u903c\u8fd1 70% \u5230 80%\uff0c\u6302\u949f\u4e00\u62e8\u56de\u771f\u5b9e\u6d41\u901d\u5c31\u65ad\u5d16\u5f0f\u8df3\u6c34\u2014\u2014\u5dee\u8ddd\u5168\u90e8\u6765\u81ea\u65f6\u95f4\u6210\u672c\u3002\u7f3a\u5c11\u65f6\u95f4\u611f\u77e5\u7684\u957f\u601d\u7ef4\u94fe\uff0c\u5728\u5b9e\u65f6\u73af\u5883\u91cc\u4f1a\u4ece\u8d44\u4ea7\u53d8\u6210\u8d1f\u503a\u3002\u8fd9\u9884\u793a\u7740\u4e0b\u4e00\u4ee3\u5b9e\u65f6\u6a21\u578b\u9700\u8981\u5728\u300c\u65f6\u95f4\u5145\u88d5\u65f6\u7ec6\u60f3\u3001\u7a97\u53e3\u5c06\u5173\u65f6\u5feb\u52a8\u3001\u65b0\u4fe1\u606f\u5230\u6765\u65f6\u6539\u8ba1\u5212\u300d\u4e4b\u95f4\u505a\u52a8\u6001\u8c03\u5ea6\uff0c\u800c\u4e0d\u662f\u4e00\u5473\u5806\u63a8\u7406\u3002<\/p>\n<p>\u4e5f\u6709\u5916\u90e8\u89c2\u5bdf\u8005\u7ed9\u51fa\u4e86\u66f4\u514b\u5236\u7684\u89e3\u8bfb\u3002Jack Clark \u5728 Import AI \u65f6\u8bc4\u4e2d\u8ba4\u4e3a\uff0c\u8fd9\u9879\u57fa\u51c6\u7684\u4ef7\u503c\u5728\u4e8e\u7cbe\u51c6\u6807\u51fa\u4e86 AI agent \u7684\u80fd\u529b\u65ad\u70b9\uff1a\u4f4e\u5206\u4efb\u52a1\u6070\u6070\u662f\u5f53\u524d\u7cfb\u7edf\u8fd8\u4e0d\u8be5\u88ab\u6258\u4ed8\u7684\u573a\u666f\uff0c\u8ba9\u6a21\u578b\u66ff\u7528\u6237\u53c2\u52a0\u5728\u7ebf\u7ade\u62cd\u6216\u6253\u5b9e\u65f6\u6e38\u620f\u4ecd\u4e3a\u65f6\u5c1a\u65e9\u3002\u8fd8\u6709\u5206\u6790\u6307\u51fa\u57fa\u51c6\u672c\u8eab\u7684\u5c40\u9650\u2014\u2014\u73af\u5883\u4ee5\u7ea6 1 \u5e27\u6bcf\u79d2\u7684\u8282\u594f\u91c7\u6837\uff0c\u53ef\u80fd\u4f4e\u4f30\u4e86\u771f\u5b9e\u4ea4\u4e92\u7684\u590d\u6742\u5ea6\u3002\u4e24\u79cd\u58f0\u97f3\u6307\u5411\u540c\u4e00\u4e2a\u5224\u65ad\uff1aRealtime-Bench \u6ca1\u6709\u5426\u5b9a\u73b0\u6709\u6a21\u578b\uff0c\u4f46\u5b83\u628a\u300c\u5b9e\u65f6\u53ef\u9760\u6027\u300d\u4ece\u5bf9\u8bdd\u6d41\u7545\u5ea6\u4e4b\u5916\u5355\u72ec\u62c9\u4e86\u51fa\u6765\uff0c\u6210\u4e3a\u4e0b\u4e00\u4ee3\u591a\u6a21\u6001\u6a21\u578b\u7ed5\u4e0d\u5f00\u7684\u5fc5\u7b54\u9898\u3002<\/p>\n<p>&#8212;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u82f1\u4f1f\u8fbe\u8054\u5408\u5f97\u514b\u8428\u65af\u5927\u5b66\u5965\u65af\u6c40\u5206\u6821\u7b49\u673a\u6784\u7684<\/p>\n","protected":false},"author":12,"featured_media":20908,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"colormag_page_container_layout":"default_layout","colormag_page_sidebar_layout":"default_layout","footnotes":""},"categories":[1],"tags":[],"class_list":["post-20909","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"acf":[],"_links":{"self":[{"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/posts\/20909","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/users\/12"}],"replies":[{"embeddable":true,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/comments?post=20909"}],"version-history":[{"count":1,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/posts\/20909\/revisions"}],"predecessor-version":[{"id":21425,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/posts\/20909\/revisions\/21425"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/media\/20908"}],"wp:attachment":[{"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/media?parent=20909"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/categories?post=20909"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dguan.com.cn\/ai\/wp-json\/wp\/v2\/tags?post=20909"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}