
{"id":67,"date":"2026-10-21T10:46:27","date_gmt":"2026-10-21T02:46:27","guid":{"rendered":"https:\/\/aicrunchx.com\/cn\/?p=67"},"modified":"2026-10-21T12:32:02","modified_gmt":"2026-10-21T04:32:02","slug":"surge-ceo-edwin-chen%e6%b7%b1%e5%ba%a6%e5%af%b9%e8%af%9d%ef%bc%9a%e5%88%ab%e8%a2%abllm-arena%e5%b8%a6%e5%81%8f%ef%bc%8c%e7%9c%9f%e6%ad%a3%e7%9a%84%e4%ba%ba%e5%b7%a5%e8%af%84%e4%bc%b0%e3%80%81rl","status":"publish","type":"post","link":"https:\/\/aicrunchx.com\/cn\/archives\/67","title":{"rendered":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3"},"content":{"rendered":"<blockquote class=\"wp-block-quote\">\n<p>\u672c\u671f\u300aUnsupervised Learning\u300b\u7531 Jacob Effron \u4e3b\u6301\uff0c\u5bf9\u8bdd\u5609\u5bbe\u662f\u4f30\u503c\u7ea6 240 \u4ebf\u7f8e\u5143\u7684\u6570\u636e\u4e0e\u6a21\u578b\u670d\u52a1\u516c\u53f8 Surge \u7684\u9996\u5e2d\u6267\u884c\u5b98 Edwin Chen\u3002Surge \u4e0e\u5168\u7403\u4e00\u7ebf\u5927\u6a21\u578b\u5b9e\u9a8c\u5ba4\u7d27\u5bc6\u5408\u4f5c\u3001\u5e2e\u52a9\u5176\u6539\u8fdb\u6a21\u578b\uff1b\u8fd9\u573a\u8fd1 48 \u5206\u949f\u7684\u957f\u8c08\u4ece&#8221;\u4e3a\u4ec0\u4e48\u4f18\u5316 LLM Arena \u662f\u5728\u4f18\u5316\u6807\u9898\u515a&#8221;\u5207\u5165\uff0c\u7cfb\u7edf\u8ba8\u8bba\u4e86\u9519\u8bef\u57fa\u51c6\u7684\u4ee3\u4ef7\u3001\u5408\u683c\u8bc4\u4f30\u5e08\u7684\u56db\u79cd\u7279\u8d28\u3001RL \u73af\u5883\u4e3a\u4f55\u6210\u4e3a\u8bad\u7ec3\u8303\u5f0f\u7684\u4e0b\u4e00\u6b65\u3001\u7845\u8c37\u7684 pivot \u6587\u5316\uff0c\u4ee5\u53ca\u4ed6\u4ece&#8221;\u4e00\u4e2a\u4e00\u7edf\u5929\u4e0b\u7684\u6a21\u578b&#8221;\u8f6c\u5411&#8221;\u6a21\u578b\u661f\u5ea7&#8221;\u7684\u89c2\u5ff5\u53d8\u5316\u3002<\/p>\n<\/blockquote>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/edwin-chen-podcast-opening.jpg\" alt=\"Edwin Chen podcast interview opening\"\/><figcaption>Podcast studio interview<\/figcaption><\/figure>\n<h2>\u4e00\u3001\u5f00\u573a\uff1a\u5750\u5728\u57fa\u5ea7\u6a21\u578b\u751f\u6001\u7b2c\u4e00\u6392\u7684\u4eba<\/h2>\n<p>\u4f5c\u4e3a Surge \u7684 CEO\uff0cEdwin Chen \u53ef\u4ee5\u8bf4\u5750\u5728\u6574\u4e2a\u57fa\u5ea7\u6a21\u578b\u751f\u6001\u7684\u7b2c\u4e00\u6392\u3002\u8fd9\u5bb6\u636e\u62a5\u9053\u4f30\u503c\u7ea6 240 \u4ebf\u7f8e\u5143\u7684\u516c\u53f8\uff0c\u4e0e\u51e0\u4e4e\u6240\u6709\u9876\u7ea7\u5b9e\u9a8c\u5ba4\u5408\u4f5c\uff0c\u6df1\u5ea6\u53c2\u4e0e\u5176\u6a21\u578b\u6539\u8fdb\u5de5\u4f5c\u3002\u5bf9\u4e00\u5bb6\u4ee5&#8221;\u4e3a AGI \u63d0\u4f9b\u6240\u9700\u4e00\u5207\u6570\u636e&#8221;\u4e3a\u4f7f\u547d\u7684\u516c\u53f8\u800c\u8a00\uff0c\u8fd9\u610f\u5473\u7740\u5b83\u6bcf\u5929\u90fd\u5728\u8fd1\u8ddd\u79bb\u89c2\u5bdf\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u5982\u4f55\u9009\u76ee\u6807\u3001\u5982\u4f55\u8bad\u7ec3\u3001\u5982\u4f55\u5ea6\u91cf\u81ea\u5df1\u7684\u6a21\u578b\u3002\u4e3b\u6301\u4eba Jacob Effron \u60f3\u4ece\u4e00\u4e2a Edwin \u53cd\u590d\u5f3a\u8c03\u7684\u89c2\u5bdf\u5f00\u59cb\uff1a\u5f53\u6a21\u578b\u6784\u5efa\u8005\u9009\u9519\u4e86\u57fa\u51c6\uff08benchmark\uff09\u53bb\u4f18\u5316\uff0c\u4f1a\u4ed8\u51fa\u600e\u6837\u7684\u4ee3\u4ef7\u3002Edwin \u7279\u522b\u70b9\u540d\u4e86 LLM Arena\uff08Alam Arena\uff09\u4ee5\u53ca\u5176\u4ed6\u4e00\u4e9b\u88ab\u8fc7\u5ea6\u8ffd\u6367\u7684\u516c\u5f00\u6392\u884c\u699c\u3002<\/p>\n<h2>\u4e8c\u3001\u4f18\u5316 LLM Arena\uff0c\u672c\u8d28\u4e0a\u662f\u5728\u4f18\u5316&#8221;\u6807\u9898\u515a&#8221;<\/h2>\n<p>Edwin \u7ed9\u51fa\u7684\u7b2c\u4e00\u4e2a\u5224\u65ad\u76f8\u5f53\u76f4\u63a5\uff1a\u5982\u679c\u4f60\u4f18\u5316 LLM Arena\uff0c\u672c\u8d28\u4e0a\u662f\u5728\u4f18\u5316\u6807\u9898\u515a\uff08clickbait\uff09\u3002\u4ed6\u63cf\u8ff0\u4e86 Arena \u7684\u4f7f\u7528\u573a\u666f\uff1a\u7528\u6237\u8fdb\u5165\u7ad9\u70b9\u3001\u53d1\u51fa\u4e00\u4e2a prompt\u3001\u5f97\u5230\u4e24\u6761\u56de\u7b54\uff0c\u7136\u540e\u6295\u7968\u54ea\u6761\u66f4\u597d\uff1b\u4f46\u7528\u6237\u5e76\u4e0d\u4f1a\u4ed4\u7ec6\u9605\u8bfb\u56de\u7b54\uff0c\u4ed6\u4eec\u5f80\u5f80\u626b\u8fc7\u4e24\u6761\u56de\u7b54\u4e00\u4e24\u79d2\uff0c\u5c31\u9009\u90a3\u4e2a&#8221;\u6700\u6253\u52a8\u81ea\u5df1&#8221;\u7684\u3002\u4ed6\u4eec\u4e0d\u4f1a\u9010\u6761\u6838\u5bf9&#8221;\u8fd9\u6761\u662f\u5426\u5b8c\u6574\u9075\u5faa\u4e86\u6211\u7684\u6307\u4ee4&#8221;&#8221;\u90a3\u6761\u662f\u5426\u5b8c\u5168\u51c6\u786e&#8221;&#8221;\u8fd9\u6761\u662f\u5426\u7814\u7a76\u5145\u5206\u3001\u8d28\u91cf\u4e0a\u4e58&#8221;\u3002\u5728\u8fd9\u79cd\u5feb\u901f\u626b\u89c6\u4e0b\uff0c\u771f\u6b63\u88ab\u4f18\u5316\u51fa\u6765\u7684\u662f\uff1a\u54ea\u6761\u56de\u7b54\u7528\u4e86\u66f4\u591a emoji\u3001\u54ea\u6761\u66f4\u5438\u5f15\u773c\u7403\u3001\u54ea\u6761\u5806\u4e86\u66f4\u591a markdown \u548c\u6807\u9898\u5c42\u7ea7\u3002\u7528\u6237\u5929\u7136\u504f\u597d\u66f4\u957f\u7684\u56de\u7b54\u2014\u2014\u66f4\u957f\u5c31\u81ea\u5e26\u4e00\u79cd&#8221;\u4e13\u4e1a\u5149\u6cfd&#8221;\uff0c\u5373\u4f7f\u5b83\u5e76\u6ca1\u6709\u88ab\u8ba4\u771f\u8bfb\u5b8c\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/bad-benchmark-optimization.jpg\" alt=\"Optimizing Arena rewards verbose clickbait\"\/><figcaption>Bad benchmark optimization<\/figcaption><\/figure>\n<p>Edwin \u4e3e\u4e86\u4e00\u4e2a\u4ed6\u6700\u8fd1\u5728\u516c\u5f00\u6570\u636e\u96c6\u4e2d\u770b\u5230\u7684\u4f8b\u5b50\u3002\u95ee\u9898\u662f&#8221;\u7ed9\u51fa 1452 \u7684\u6240\u6709\u9664\u6570\uff081452 \u4ee5\u4e0b\u3001\u80fd\u6574\u9664\u5b83\u7684\u6570\uff09&#8221;\uff0c\u6b63\u786e\u7b54\u6848\u5e94\u5305\u542b 1\u30012\u30013\u30014\u30016 \u7b49\uff1b\u5176\u4e2d\u4e00\u6761\u56de\u7b54\u8bf4&#8221;1452 \u53ea\u6709\u4e00\u4e2a\u9664\u6570\uff0c\u5c31\u662f 1&#8243;\uff0c\u5b8c\u5168\u9519\u8bef\uff0c\u53e6\u4e00\u6761\u5219\u7b54\u5f97\u5b8c\u5168\u6b63\u786e\u3002\u4f46\u7ffb\u770b\u7528\u6237\u6295\u7968\u7ed3\u679c\uff0c\u88ab\u504f\u597d\u7684\u6070\u6070\u662f\u90a3\u6761\u5b8c\u5168\u9519\u8bef\u7684\u56de\u7b54\u3002\u4ed6\u6307\u51fa\uff0c\u8fd9\u53ea\u662f\u4e00\u4e2a\u76f8\u5f53\u7b80\u5355\u7684\u6570\u5b66\u95ee\u9898\uff1b\u6362\u6210\u66f4\u9ad8\u7ea7\u7684 prompt\uff0c\u666e\u901a\u7528\u6237\u66f4\u4e0d\u4f1a\u81ea\u5df1\u53bb\u67e5\u8bc1\u3001\u53bb\u4e8b\u5b9e\u6838\u67e5\u2014\u2014\u8fd9\u4e0d\u662f\u7528\u6237\u4f1a\u505a\u7684\u4e8b\u3002\u7c7b\u4f3c\u5730\uff0c\u4ed6\u63d0\u5230\u4e00\u5e74\u524d\u6709\u4e00\u9879\u5173\u4e8e ChatGPT \u533b\u7597\u56de\u7b54\u7684\u7814\u7a76\uff1aAI \u56de\u7b54\u8bc4\u5206\u8fdc\u9ad8\u4e8e\u533b\u751f\u56de\u7b54\uff0c\u62c6\u89e3\u539f\u56e0\u540e\u53d1\u73b0\uff0c\u51e0\u4e4e\u53ea\u662f\u56e0\u4e3a\u56de\u7b54\u66f4\u957f\u3002<\/p>\n<h2>\u4e09\u3001\u9519\u8bef\u76ee\u6807\u53e0\u52a0\u9519\u8bef\u6570\u636e\u4e0e\u7f3a\u4e4f\u5ea6\u91cf\uff1a\u6a21\u578b\u5012\u9000\u516d\u4e2a\u6708\u800c\u4e0d\u81ea\u77e5<\/h2>\n<p>Edwin \u628a\u95ee\u9898\u8fdb\u4e00\u6b65\u5f52\u7eb3\u4e3a\u4e00\u4e2a\u7ec4\u5408\uff1a\u5f53\u4f60\u540c\u65f6\u5728\u4f18\u5316\u9519\u8bef\u7684\u5bf9\u8c61\u3001\u7528\u9519\u8bef\u7684\u987a\u5e8f\u4f18\u5316\u3001\u4f18\u5316\u9519\u8bef\u7684\u76ee\u6807\u51fd\u6570\uff0c\u518d\u52a0\u4e0a\u6ca1\u6709\u5efa\u7acb\u6b63\u786e\u7684\u5ea6\u91cf\uff0c\u9ebb\u70e6\u5c31\u4f1a\u51fa\u73b0\u3002\u4eca\u5e74\u65e9\u4e9b\u65f6\u5019\uff0cSurge \u4e0e\u4e00\u4e2a\u65b0\u56e2\u961f\u5408\u4f5c\uff0c\u7814\u7a76\u5458\u4eec\u5df2\u7ecf\u6000\u7591\u81ea\u5df1\u7684\u6a21\u578b\u5728\u53d8\u5dee\uff0c\u5374\u62ff\u4e0d\u51fa\u5b9a\u91cf\u8bc1\u636e\uff0c\u56e0\u4e3a\u4ed6\u4eec\u6839\u672c\u6ca1\u6709\u5408\u9002\u7684\u5ea6\u91cf\u624b\u6bb5\u3002Surge \u6df1\u5165\u6392\u67e5\u540e\u53d1\u73b0\uff0c\u8fc7\u53bb 6 \u5230 12 \u4e2a\u6708\uff0c\u4ed6\u4eec\u7684\u6a21\u578b\u5b9e\u9645\u4e0a\u5df2\u7ecf\u5728\u5012\u9000\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/misaligned-training-pitfalls.jpg\" alt=\"Misaligned optimization hides regressions\"\/><figcaption>Training misalignment pitfalls<\/figcaption><\/figure>\n<p>\u6839\u6e90\u51fa\u5728\u6570\u636e\u4e0a\uff1a\u6240\u8c13&#8221;\u8d44\u6df1\u7f16\u7801\u8005&#8221;\u4ea7\u51fa\u7684\u8bad\u7ec3\u6570\u636e\u3001\u8bc4\u5ba1\u8005\u7ed9\u51fa\u7684\u53cd\u9988\uff0c\u5e76\u6ca1\u6709\u771f\u6b63\u6267\u884c\u4ee3\u7801\u3001\u4ed4\u7ec6\u6838\u67e5\u4ee3\u7801\u662f\u5426\u6b63\u786e\u3002\u4ed6\u4eec\u7ed9\u51fa\u7684\u8bad\u7ec3\u6570\u636e\u5145\u6ee1\u534e\u4e3d\u8f9e\u85fb\u548c\u5b8f\u5927\u5ba3\u79f0\u2014\u2014&#8221;\u6211\u4e3a\u4f60\u5199\u51fa\u4e86\u8fd9\u4e2a\u80fd\u5b9e\u73b0 A\u3001B\u3001C \u7684\u60ca\u8273\u7a0b\u5e8f&#8221;\uff0c\u5374\u6ca1\u6709\u82b1\u65f6\u95f4\u771f\u6b63\u8fd0\u884c\u3002\u800c\u6267\u884c\u4ee3\u7801\u5f80\u5f80\u5e76\u4e0d\u8f7b\u677e\uff1a\u8981\u5b89\u88c5\u5404\u79cd\u5e93\u3001\u914d\u9f50\u57fa\u7840\u8bbe\u65bd\u3001\u771f\u6b63\u7406\u89e3\u8fd9\u95e8\u8bed\u8a00\u3002\u7ed3\u679c\u8fd9\u4e9b\u4ee3\u7801\u8981\u4e48\u5b8c\u5168\u9519\u8bef\uff0c\u8981\u4e48\u57cb\u7740\u8bb8\u591a\u65e5\u540e\u624d\u66b4\u9732\u7684\u7ec6\u5fae bug\u3002\u5728\u6574\u4e2a\u884c\u4e1a\u90fd\u5728\u524d\u8fdb\u3001\u7f16\u7801\u53c8\u662f\u6700\u70ed\u8d5b\u9053\u7684\u80cc\u666f\u4e0b\uff0c\u8fd9\u652f\u56e2\u961f\u5374\u56e0\u4e3a\u6570\u636e\u548c\u5ea6\u91cf\u53cc\u53cc\u7f3a\u4f4d\uff0c\u5b9e\u9645\u4e0a\u5728\u505a\u8d1f\u5411\u8fdb\u6b65\u3002Edwin \u8ba4\u4e3a\uff0c\u771f\u6b63\u503c\u5f97\u8b66\u60d5\u7684\u662f\uff1a\u884c\u4e1a\u91cc\u76f8\u5f53\u591a\u7684\u4eba\u6ca1\u6709\u8ba4\u771f\u5173\u6ce8\u81ea\u5df1\u6536\u5230\u7684\u6570\u636e\u8d28\u91cf\uff0c\u4e5f\u6ca1\u6709\u5728\u5ea6\u91cf\u6b63\u786e\u7684\u4e1c\u897f\u3002<\/p>\n<h2>\u56db\u3001\u5b66\u672f\u57fa\u51c6\u8fc7\u62df\u5408\uff1aSAT \u586b\u9e2d\u5f0f\u7c7b\u6bd4\uff0c\u5206\u6570\u6da8\u800c\u771f\u5b9e\u80fd\u529b\u964d<\/h2>\n<p>\u5386\u53f2\u4e0a\uff0c\u6240\u6709\u9ad8\u5ea6\u91cd\u89c6\u57fa\u51c6\u7684\u516c\u53f8\u90fd\u76ef\u7740\u7814\u7a76\u793e\u533a\u6784\u5efa\u7684\u90a3\u4e9b\u5b66\u672f\u57fa\u51c6\u3002Edwin \u6307\u51fa\uff0c\u6a21\u578b\u672c\u8d28\u4e0a\u975e\u5e38\u64c5\u957f&#8221;\u722c\u5c71&#8221;\uff0c\u975e\u5e38\u9002\u5408\u5177\u4f53\u3001\u72ed\u7a84\u3001\u5b9a\u4e49\u660e\u786e\u7684\u76ee\u6807\u51fd\u6570\uff0c\u4e8e\u662f\u5f88\u5bb9\u6613\u5728\u57fa\u51c6\u4e0a\u53d6\u5f97\u5927\u91cf\u8fdb\u6b65\u2014\u2014\u6709\u65f6\u751a\u81f3\u662f&#8221;\u5047\u8fdb\u6b65&#8221;\uff0c\u6bd4\u5982\u57fa\u51c6\u6570\u636e\u672c\u6765\u5c31\u6df7\u8fdb\u4e86\u8bad\u7ec3\u96c6\u800c\u65e0\u4eba\u5bdf\u89c9\uff1b\u5373\u4fbf\u4e0d\u662f\u5047\u8fdb\u6b65\uff0c\u6a21\u578b\u5728\u8fd9\u4e9b\u72ed\u7a84\u95ee\u9898\u4e0a\u53d8\u597d\u7684\u540c\u65f6\uff0c\u5728\u66f4\u771f\u5b9e\u4e16\u754c\u7684\u95ee\u9898\u4e0a\u5176\u5b9e\u5728\u53d8\u5dee\uff0c\u800c\u56e2\u961f\u56e0\u4e3a\u57fa\u51c6\u4e4b\u5916\u6ca1\u6709\u5ea6\u91cf\u800c\u65e0\u4ece\u77e5\u6653\u3002<\/p>\n<p>\u4ed6\u7528 SAT \u586b\u9e2d\u4f5c\u7c7b\u6bd4\uff1a\u4e00\u4e2a\u9ad8\u4e2d\u751f\u53ef\u4ee5\u82b1\u51e0\u767e\u4e0a\u5343\u5c0f\u65f6\u4e13\u95e8\u5237 SAT\uff0cSAT \u53ea\u8003\u9605\u8bfb\u7406\u89e3\u3001\u7c7b\u6bd4\u3001\u8bcd\u6c47\u8fd9\u4e00\u5957\u72ed\u7a84\u9898\u76ee\uff0c\u4f46\u5b83\u5e76\u4e0d\u6d4b\u91cf\u4f60\u5199\u597d\u6587\u7ae0\u7684\u80fd\u529b\u3001\u89e3\u51b3\u590d\u6742\u73b0\u5b9e\u95ee\u9898\u7684\u80fd\u529b\uff0c\u4ee5\u53ca SAT \u4e4b\u5916\u7684\u8bb8\u591a\u80fd\u529b\u3002Surge \u81ea\u5df1\u628a\u73a9\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u7ed9\u51fa\u7684\u6a21\u578b\u65f6\u7ecf\u5e38\u53d1\u73b0\uff0c\u516c\u5f00\u57fa\u51c6\u5206\u6570\u5f88\u6f02\u4eae\uff0c\u5b9e\u9645\u6a21\u578b\u5374&#8221;\u5dee\u4e86\u4e00\u5927\u622a&#8221;\u3002\u5f53\u67d0\u4e2a\u6a21\u578b\u7a81\u7136\u8d28\u91cf\u4e0b\u6ed1\uff0c\u56e2\u961f\u5f80\u5f80\u5766\u8a00\uff1a\u4e3a\u4e86\u5237\u67d0\u4e2a\u57fa\u51c6\uff0c\u4ed6\u4eec\u5806\u4e86\u5927\u91cf\u5408\u6210\u6570\u636e\u51b2\u5206\uff0c\u5374\u6ca1\u610f\u8bc6\u5230\u5176\u4ed6\u540e\u679c\u3002\u8fd9\u4e5f\u662f\u4e3a\u4ec0\u4e48\u4e00\u7ebf\u5b9e\u9a8c\u5ba4\u9010\u6e10\u8ba4\u8bc6\u5230\uff1a\u552f\u4e00\u53ef\u9760\u7684\u5ea6\u91cf\u65b9\u5f0f\uff0c\u662f\u8fd0\u884c\u89c4\u8303\u7684\u4eba\u5de5\u8bc4\u4f30\uff08human evals\uff09\u3002<\/p>\n<h2>\u4e94\u3001\u552f\u4e00\u53ef\u9760\u7684\u5ea6\u91cf\uff1a\u89c4\u8303\u800c\u4e25\u8c28\u7684\u4eba\u5de5\u8bc4\u4f30<\/h2>\n<p>\u5728\u89c4\u8303\u7684\u4eba\u5de5\u8bc4\u4f30\u4e2d\uff0c\u7531\u9ad8\u8d28\u91cf\u3001\u52e4\u52c9\u3001\u771f\u6b63\u5173\u6ce8\u56de\u7b54\u5185\u5bb9\u7684\u4eba\uff0c\u540c\u65f6\u5177\u5907\u6210\u719f\u5ea6\u548c\u54c1\u5473\u53bb\u5224\u65ad\uff1a\u8fd9\u662f\u4e0d\u662f\u5b9e\u9a8c\u5ba4\u60f3\u8981\u7684\u4eba\u683c\u3001\u662f\u4e0d\u662f\u5b83\u5e0c\u671b\u6a21\u578b\u6a21\u4eff\u7684\u98ce\u683c\u3002\u5b83\u672c\u8d28\u4e0a\u662f\u5728\u590d\u523b\u771f\u5b9e\u7528\u6237\u3001\u771f\u5b9e\u4e16\u754c\u7684\u4f53\u9a8c\u2014\u2014\u628a\u771f\u5b9e\u4e16\u754c\u7684\u591a\u6837\u6027\u3001\u590d\u6742\u6027\u3001\u6742\u4e71\u6027\u90fd\u7eb3\u5165\u8fdb\u6765\uff0c\u800c\u4e0d\u53d7\u5b98\u65b9\u57fa\u51c6\u90a3\u79cd\u4eba\u4e3a\u7ea6\u675f\uff1b\u5e76\u7531\u7ecf\u9a8c\u4e30\u5bcc\u3001\u503c\u5f97\u4fe1\u4efb\u7684\u8bc4\u4f30\u8005\u6267\u884c\uff0c\u800c\u975e\u6f2b\u4e0d\u7ecf\u5fc3\u7684\u533f\u540d\u7528\u6237\u3002\u8fd9\u5957\u4e25\u8c28\u7684\u4eba\u5de5\u8bc4\u4f30\u6d41\u7a0b\uff0c\u5df2\u7ecf\u6210\u4e3a\u6240\u6709\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u516c\u8ba4\u7684\u9ec4\u91d1\u6807\u51c6\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/rigorous-human-evals.jpg\" alt=\"Rigorous human evaluation gold standard\"\/><figcaption>Human evaluation standard<\/figcaption><\/figure>\n<h2>\u516d\u3001\u5408\u683c\u8bc4\u4f30\u5e08\u7684\u56db\u79cd\u7279\u8d28\uff1a\u4e13\u4e1a\u3001\u54c1\u5473\u3001\u521b\u9020\u529b\u3001\u9075\u5faa\u6307\u4ee4<\/h2>\n<p>\u5f53\u88ab\u8ffd\u95ee&#8221;\u4ec0\u4e48\u6837\u7684\u8bc4\u4f30\u5e08\u6709\u54c1\u5473&#8221;\u65f6\uff0cEdwin \u5f52\u7eb3\u51fa\u56db\u70b9\u3002\u7b2c\u4e00\u662f\u4e13\u4e1a\u77e5\u8bc6\u672c\u8eab\uff1aSurge \u8bb8\u591a\u8bc4\u4f30\u96c6\u975e\u5e38\u9ad8\u7ea7\uff0c\u53ef\u80fd\u5728\u6d4b\u91cf\u6a21\u578b\u505a\u4ee3\u6570\u62d3\u6251\u7814\u7a76\u7684\u80fd\u529b\uff0c\u6216\u4f7f\u7528 PyTorch \u7684\u80fd\u529b\uff0c\u8bc4\u4f30\u8005\u9996\u5148\u5fc5\u987b\u662f\u5728\u8be5\u9886\u57df\u6709\u6df1\u539a\u4e13\u957f\u3001\u8db3\u591f\u7528\u5fc3\u7684\u4eba\u3002\u7b2c\u4e8c\u662f\u6210\u719f\u5ea6\u4e0e\u54c1\u5473\uff1a\u5199\u4ee3\u7801\u4f60\u4e0d\u53ea\u8981\u6b63\u786e\u6027\uff0c\u8fd8\u8981\u770b\u4ee3\u7801\u662f\u5426\u6574\u6d01\u3001\u8bbe\u8ba1\u662f\u5426\u826f\u597d\uff1b\u5199\u6587\u7ae0\u4f60\u4e0d\u53ea\u8981&#8221;\u5199\u51fa\u6765&#8221;\uff0c\u8fd8\u8981\u770b\u662f\u5426\u63d0\u51fa\u4e86\u65b0\u60f3\u6cd5\u3001\u884c\u6587\u662f\u5426\u51fa\u8272\u3001\u662f\u5426\u4e0d\u50cf AI \u5783\u573e\uff08AI slop\uff09\u3002\u7b2c\u4e09\u662f\u521b\u9020\u529b\uff1a\u4eba\u4eec\u5f80\u5f80\u4f4e\u4f30 prompt \u7684\u91cd\u8981\u6027\uff0c\u8981\u6b63\u786e\u5ea6\u91cf\u6a21\u578b\uff0cprompt \u5fc5\u987b\u8986\u76d6\u4f60\u5e0c\u671b\u6a21\u578b\u64c5\u957f\u7684\u6574\u4e2a\u80fd\u529b\u5206\u5e03\uff1b\u5982\u679c\u4e00\u5343\u6761 prompt \u90fd\u7528\u540c\u4e00\u79cd\u53e5\u5f0f\uff0c\u800c\u4e0d\u662f\u8986\u76d6\u771f\u5b9e\u4e16\u754c\u7528\u6237\u4ea4\u4e92\u7684\u957f\u5c3e\uff0c\u90a3\u5c31\u8fdc\u8fdc\u4e0d\u591f\u3002\u521b\u9020\u529b\u5730\u8bbe\u8ba1 prompt \u51fa\u4eba\u610f\u6599\u5730\u96be\u3002\u7b2c\u56db\u662f\u9075\u5faa\u6307\u4ee4\u7684\u80fd\u529b\uff1a\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u5e38\u6709\u975e\u5e38\u5177\u4f53\u7684\u6807\u51c6\u3001\u98ce\u683c\u6307\u5357\u3001\u4eba\u683c\u8981\u6c42\uff0c\u6216\u5bf9\u67d0\u4e9b\u6807\u51c6\u4f18\u5148\u4e8e\u5176\u4ed6\u6807\u51c6\u7684\u590d\u6742\u89c4\u5b9a\uff0c\u8bc4\u4f30\u8005\u8981\u80fd\u4e25\u683c\u9075\u5faa\u8fd9\u4e9b\u590d\u6742\u6307\u4ee4\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/good-evaluator-traits.jpg\" alt=\"Four traits of a good evaluator\"\/><figcaption>Good evaluator traits<\/figcaption><\/figure>\n<h2>\u4e03\u3001RL \u73af\u5883\uff1aSFT\u3001RLHF\u3001\u9a8c\u8bc1\u5668\u4e4b\u540e\u7684\u4e0b\u4e00\u6b65<\/h2>\n<p>\u8c08\u5230\u5f53\u4e0b\u5927\u70ed\u7684 RL \u73af\u5883\u53ca\u5176\u5956\u52b1\u6a21\u578b\uff0cEdwin \u8ba4\u4e3a\u73af\u5883\u662f\u8bad\u7ec3\u8303\u5f0f\u6f14\u8fdb\u7684\u81ea\u7136\u4e0b\u4e00\u6b65\uff1a\u5386\u53f2\u4e0a\u5927\u91cf\u5de5\u4f5c\u6295\u5165 SFT\uff0c\u7136\u540e\u662f RLHF\uff0c\u518d\u662f\u9a8c\u8bc1\u5668\uff0c\u800c\u73af\u5883\u53ea\u662f\u8fd9\u6761\u8fdb\u6b65\u9636\u68af\u4e0a\u7684\u4e0b\u4e00\u6b65\uff0c\u672a\u6765\u8fd8\u4f1a\u6709\u522b\u7684\u6b65\u9aa4\u3002\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cSurge \u5df2\u7ecf\u5728\u81ea\u5df1\u7684\u73af\u5883\u4e0a\u5de5\u4f5c\u4e86\u4e00\u4e24\u5e74\uff0c\u800c\u6574\u4e2a\u884c\u4e1a\u6700\u8fd1\u624d\u5f00\u59cb\u8ddf\u8fdb\u2014\u2014\u4ed6\u4eec\u4e0e Meta \u7684 agents \u56e2\u961f\u5408\u4f5c\u73af\u5883\u5df2\u7ecf\u8d85\u8fc7\u4e00\u5e74\uff0c\u8fd9\u652f\u56e2\u961f\u6b63\u662f Gaia \u57fa\u51c6\u7684\u7f14\u9020\u8005\uff0c\u8fd1\u671f\u4e5f\u5f00\u6e90\u4e86\u81ea\u5df1\u7684 agent \u7814\u7a76\u73af\u5883\u5e73\u53f0\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/rl-environments-next-step.jpg\" alt=\"RL environments next training paradigm step\"\/><figcaption>RL environment next step<\/figcaption><\/figure>\n<p>\u8981\u6784\u5efa\u8fd9\u6837\u7684\u73af\u5883\uff0c\u9700\u8981\u51e0\u7c7b\u4e1c\u897f\uff1a\u9996\u5148\u662f\u5c3d\u53ef\u80fd\u62df\u771f\u3001\u590d\u6742\u7684\u6a21\u62df\u4e16\u754c\uff0c\u91cc\u9762\u586b\u5145\u4eba\u3001\u4f01\u4e1a\u3001\u5de5\u5177\u4ee5\u53ca\u4eba\u4e0e\u4eba\u4e4b\u95f4\u7684\u4e92\u52a8\u2014\u2014\u6d88\u606f\u3001Slack\u3001\u90ae\u4ef6\u3001\u65e5\u5386\u4e8b\u4ef6\u7b49\uff0c\u90fd\u8981\u4ee5\u6709\u8da3\u800c\u590d\u6742\u7684\u65b9\u5f0f\u590d\u523b\u771f\u5b9e\u4e16\u754c\uff1b\u5176\u6b21\u662f\u652f\u6491\u6a21\u578b\u8fd0\u884c\u7684\u5e95\u5c42\u5de5\u5177\u4e0e\u57fa\u7840\u8bbe\u65bd\uff0c\u5305\u62ec\u6a21\u578b\u8bbf\u95ee\u7684 MCP \u670d\u52a1\u5668\u3001\u6d4f\u89c8\u5668\u3001\u6a21\u578b\u6267\u884c\u7684\u4ee3\u7801\u73af\u5883\uff1b\u7136\u540e\u8981\u8bbe\u8ba1\u8003\u9a8c\u524d\u6cbf\u6a21\u578b\u6781\u9650\u7684\u4efb\u52a1\u5e76\u6d4b\u8bd5\u8868\u73b0\uff1b\u8fd8\u6709\u4e00\u5927\u5757\u662f\u5ea6\u91cf\u4e0e\u5185\u7701\uff1a\u4e00\u65e6\u53d1\u73b0\u6a21\u578b\u5931\u8d25\uff0c\u5c31\u8981\u6df1\u6316\u5931\u8d25\u539f\u56e0\u3002<\/p>\n<h2>\u516b\u3001\u5173\u6ce8\u6a21\u578b\u8f68\u8ff9\uff0c\u8b66\u60d5\u5956\u52b1\u9ed1\u5ba2<\/h2>\n<p>Edwin \u63d0\u51fa\u4e00\u4e2a\u5bb9\u6613\u88ab\u5ffd\u89c6\u7684\u89c6\u89d2\uff1a\u5fc5\u987b\u5173\u6ce8\u6a21\u578b\u7684\u884c\u52a8\u8f68\u8ff9\uff08trajectories\uff09\uff0c\u7406\u89e3\u5b83\u4eec\u4e3a\u4f55\u6210\u529f\u3001\u4e3a\u4f55\u5931\u8d25\u3002\u4eba\u4eec\u5f80\u5f80\u4f4e\u4f30\u4e86\u6a21\u578b&#8221;\u5956\u52b1\u9ed1\u5ba2&#8221;\uff08reward hacking\uff09\u8d70\u5230\u6b63\u786e\u7b54\u6848\u7684\u7a0b\u5ea6\uff0c\u4ed6\u89c1\u8fc7\u8bb8\u591a\u6ed1\u7a3d\u53c8\u75af\u72c2\u7684\u4f5c\u5f0a\u6848\u4f8b\uff1b\u540c\u65f6\u4e5f\u4f4e\u4f30\u4e86\u6a21\u578b\u5931\u8d25\u7684\u591a\u79cd\u955c\u50cf\u65b9\u5f0f\uff0c\u4ee5\u53ca\u8fd9\u4e9b\u5931\u8d25\u66b4\u9732\u4e86\u6a21\u578b\u5e95\u5c42\u80fd\u529b\u7684\u4ec0\u4e48\u95ee\u9898\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/model-trajectories-hacking.jpg\" alt=\"Track model trajectories reward hacking\"\/><figcaption>Trajectories and hacking risk<\/figcaption><\/figure>\n<p>\u5b9e\u8df5\u4e2d\u5e38\u89c1\u7684\u8bef\u533a\u662f\uff1a\u53ea\u7ed9\u4e00\u4e2a\u6700\u7ec8\u5956\u52b1\uff0c\u4ee5\u4e3a\u5728\u6700\u7ec8\u5956\u52b1\u4e0a\u8bad\u7ec3\u5c31\u80fd\u89e3\u51b3\u4e00\u5207\u3002\u4f46\u6a21\u578b\u4f1a\u4ee5\u5404\u79cd\u53e4\u602a\u65b9\u5f0f\u504f\u79bb\uff0c\u6216\u5728\u4e0d\u540c\u5e95\u5c42\u80fd\u529b\u4e0a\u8868\u73b0\u51fa\u4e0d\u540c\u7c7b\u578b\u7684\u667a\u80fd\uff1b\u5982\u679c\u4e0d\u628a\u5e95\u5c42\u80fd\u529b\u592f\u5b9e\uff0c\u6a21\u578b\u77ed\u671f\u770b\u4f3c\u8868\u73b0\u826f\u597d\uff0c\u957f\u671f\u5fc5\u7136\u66b4\u9732\u51fa\u5927\u91cf\u95ee\u9898\u3002\u8fd9\u4e0e\u57fa\u51c6\u9ed1\u5ba2\u4e00\u8109\u76f8\u627f\uff1a\u5927\u5bb6\u76ef\u7740\u6570\u5b57\u4e0a\u6da8\u4ee5\u4e3a\u5728\u8fdb\u6b65\uff0c\u5e95\u5c42\u6a21\u578b\u5374\u6ca1\u6709\u5b9e\u8d28\u6027\u5730\u53d8\u5f97\u66f4\u806a\u660e\u3002<\/p>\n<h2>\u4e5d\u3001\u7845\u8c37 pivot \u6587\u5316\u4e4b\u5f0a\uff0c\u4e0e Surge \u7684\u4e09\u9879\u7ade\u4e89\u529b<\/h2>\n<p>\u6700\u8fd1\u4e00\u6279\u65b0\u521b\u516c\u53f8\u6d8c\u5165 RL \u73af\u5883\u8d5b\u9053\uff0c\u5149\u662f YC \u7cfb\u5c31\u7ea6\u6709\u4e09\u5341\u5bb6&#8221;\u73af\u5883\u5373\u670d\u52a1&#8221;\u516c\u53f8\u3002Edwin \u76f4\u8a00\u7845\u8c37\u6700\u8ba9\u4ed6\u6293\u72c2\u7684\u662f pivot \u6587\u5316\uff1a\u4eba\u4eec\u4e0d\u65ad\u8f6c\u5411\u5f53\u4e0b\u6700\u70ed\u7684\u8bdd\u9898\u6216\u80fd\u63a8\u9ad8\u4f30\u503c\u7684\u65b9\u5411\uff0c\u800c\u4e0d\u662f\u5efa\u8bbe\u81ea\u5df1\u771f\u6b63\u76f8\u4fe1\u7684\u4e1c\u897f\u3002\u4ed6\u751a\u81f3\u8bbd\u523a\u8bf4\uff0c\u7845\u8c37\u4e00\u9762\u4e0d\u9f7f\u534e\u5c14\u8857&#8221;\u8ffd\u9010\u91d1\u94b1&#8221;\uff0c\u81ea\u5df1\u5374\u5728\u8ffd\u9010\u540c\u6837\u7684\u4e1c\u897f\u2014\u2014\u8ffd\u9010\u4f30\u503c\u3001\u8ffd\u9010 VC\uff0cpivot \u5f80\u5f80\u53ea\u662f\u4e3a\u4e86\u6309 YC \u7684\u6307\u5f15\u505a\u5230\u4ea7\u54c1\u5e02\u573a\u5951\u5408\u3001\u597d\u5411\u6295\u8d44\u4eba\u5c55\u793a\u6536\u5165\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg\" alt=\"Critique of Silicon Valley pivot culture\"\/><figcaption>SV pivot culture critique<\/figcaption><\/figure>\n<p>\u8c08\u5230\u4e3a\u4f55\u8001\u724c\u73a9\u5bb6\u80fd\u4e0e\u4e09\u5341\u5bb6\u65b0\u8fdb\u5165\u8005\u7ade\u4e89\uff0cEdwin \u7ed9\u51fa\u4e09\u70b9\uff1a\u5176\u4e00\uff0c\u73af\u5883\u53ea\u662f&#8221;\u4e3a AGI \u63d0\u4f9b\u6240\u9700\u6570\u636e&#8221;\u8fd9\u4e00\u6838\u5fc3\u4f7f\u547d\u7684\u5f53\u524d\u8fed\u4ee3\uff0c\u4e0e\u516c\u53f8\u6839\u672c\u8bba\u70b9\u5929\u7136\u4e00\u81f4\uff1b\u5176\u4e8c\uff0c\u73af\u5883\u9700\u8981\u5927\u91cf\u5de5\u5177\u94fe\u2014\u2014\u521b\u5efa\u5de5\u5177\u3001\u8fd0\u884c\u6a21\u578b\u3001\u5ea6\u91cf\u3001\u5206\u6790\uff0c\u8fd9\u4e0e\u8fc7\u53bb\u505a SFT \u6570\u636e\u6240\u9700\u7684\u5de5\u5177\u5e76\u65e0\u672c\u8d28\u4e0d\u540c\uff0c\u800c Surge \u59cb\u7ec8\u662f\u4e00\u5bb6\u6280\u672f\u516c\u53f8\uff0c\u4e0d\u50cf\u5f88\u591a\u540c\u7c7b\u516c\u53f8\u53ea\u662f\u4eba\u529b\u6d3e\u9063\u673a\u6784\uff1b\u5176\u4e09\uff0c\u73af\u5883\u672c\u8d28\u662f\u8981\u4ea7\u51fa\u4e30\u5bcc\u3001\u590d\u6742\u3001\u6709\u521b\u9020\u529b\u7684\u6570\u636e\uff0c\u800c\u8fd9\u53ea\u80fd\u9760\u4eba\u5b8c\u6210\uff0c\u6b63\u5982 SWE-bench \u7531\u4eba\u521b\u5efa PR\u3001\u518d\u7531\u4eba\u6e05\u6d17\u9a8c\u8bc1\uff0c\u73af\u5883\u4e5f\u662f\u4e00\u4e2a\u9700\u8981\u6280\u672f\u652f\u6491\u7684\u4eba\u672c\u6570\u636e\u95ee\u9898\u3002<\/p>\n<h2>\u5341\u3001\u552f\u624d\u662f\u4e3e\uff1a\u5b66\u5386\u4e0d\u7b49\u4e8e\u8d28\u91cf\uff0c\u6570\u767e\u4e07\u4fe1\u53f7\u9a71\u52a8\u664b\u5347<\/h2>\n<p>\u9488\u5bf9&#8221;\u662f\u5426\u628a\u8d28\u91cf\u7b49\u540c\u4e8e\u5b66\u5386&#8221;\u7684\u7591\u95ee\uff0cEdwin \u4e3e\u4e86\u6d77\u660e\u5a01\u7684\u4f8b\u5b50\uff1a\u4ed6\u6ca1\u6709\u4e00\u5806\u535a\u58eb\u5934\u8854\uff0c\u751a\u81f3\u672a\u5fc5\u5b8c\u6210\u5927\u5b66\u5b66\u4e1a\uff0c\u800c Surge \u8981\u7684\u662f\u6bcf\u4e2a\u5c3a\u5ea6\u4e0a\u4e16\u754c\u4e0a\u6700\u9876\u5c16\u7684\u4eba\uff0c\u4e0e\u5b66\u5386\u65e0\u5173\u3002\u4ed6\u7c7b\u6bd4 Google\uff1a\u62db\u4eba\u3001\u664b\u5347\u90fd\u4e0d\u770b\u7b80\u5386\u4e0a\u7684\u5b66\u6821\u6216\u5b66\u4f4d\uff0c\u800c\u770b\u5b9e\u9645\u4ea7\u51fa\u3002Surge \u7684\u6280\u672f\u5e73\u53f0\u6bcf\u5929\u4e3a\u6bcf\u4f4d\u5de5\u4f5c\u8005\u91c7\u96c6\u6570\u767e\u4e07\u6761\u8868\u73b0\u4fe1\u53f7\uff0c\u636e\u6b64\u5ea6\u91cf\u4ed6\u4eec\u4ea7\u51fa\u7684\u6570\u636e\u8d28\u91cf\uff0c\u662f\u4ed6\u80fd\u60f3\u8c61\u5230\u6700&#8221;\u552f\u624d\u662f\u4e3e&#8221;\u7684\u673a\u5236\u2014\u2014\u4e0d\u662f\u56e0\u4e3a\u4f60\u6709\u54c8\u4f5b\u5b66\u4f4d\u5c31\u664b\u5347\uff0c\u800c\u662f\u6309\u4f60\u5b9e\u9645\u505a\u7684\u4e8b\u664b\u5347\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/meritocratic-data-curation.jpg\" alt=\"Meritocratic data curation platform\"\/><figcaption>Meritocratic data system<\/figcaption><\/figure>\n<p>\u5b66\u5386\u4e4b\u6240\u4ee5\u4e0d\u591f\u7528\uff0c\u6709\u4e24\u4e2a\u539f\u56e0\uff1a\u4e00\u662f\u4f18\u79c0\u7684 MIT \u8ba1\u7b97\u673a\u6bd5\u4e1a\u751f\u672a\u5fc5\u613f\u610f\u8ba4\u771f\u751f\u4ea7\u8bad\u7ec3\u6570\u636e\uff0c\u53cd\u800c\u53ef\u80fd\u88ab\u5bf9\u6297\u6027\u653b\u51fb\u3001\u6b3a\u9a97\u7cfb\u7edf\u5438\u5f15\uff0c\u94bb\u7a7a\u5b50\u4f5c\u5f0a\uff1b\u4e8c\u662f\u5373\u4fbf MIT \u51fa\u8eab\uff0c\u4ed6\u9762\u8bd5\u8fc7\u7684\u4eba\u91cc\u4e00\u534a\u751a\u81f3\u4e0d\u4f1a\u5199\u4ee3\u7801\uff0c\u6559\u79d1\u4e66\u4e0a\u8bfb\u61c2\u4e0e\u5728\u8857\u5934\u771f\u6b63\u6709\u667a\u6167\u5730\u6267\u884c\u4e4b\u95f4\u5dee\u8ddd\u5de8\u5927\u2014\u2014\u8fd9\u53c8\u547c\u5e94\u4e86&#8221;\u57fa\u51c6\u8868\u73b0 vs \u73b0\u5b9e\u8868\u73b0&#8221;\u7684\u8001\u95ee\u9898\u3002\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u957f\u671f\u9762\u4e34\u7684\u4e00\u4e2a\u96be\u9898\uff0c\u6b63\u662f\u6a21\u578b&#8221;\u8fc7\u4e8e\u6559\u79d1\u4e66\u5f0f\u806a\u660e&#8221;\uff0c\u5374\u7f3a\u4e4f\u73b0\u5b9e\u4e16\u754c\u7684\u8857\u5934\u667a\u6167\u3002<\/p>\n<h2>\u5341\u4e00\u3001Meta \u6536\u8d2d Scale \u7684\u5f71\u54cd\uff0c\u4e0e\u7814\u7a76\u5b9e\u9a8c\u5ba4\u5f0f\u62db\u8058<\/h2>\n<p>\u8c08\u53ca Meta \u6536\u8d2d\u5173\u7cfb\u7d27\u5bc6\u7684\u4f9b\u5e94\u5546 Scale\uff0cEdwin \u8ba4\u4e3a\u8fd9\u5bf9 Surge \u53cd\u800c\u662f\u597d\u4e8b\uff1a\u5728\u6b64\u4e4b\u524d\u4ed6\u4eec\u76f8\u5f53\u4f4e\u8c03\uff0c\u9876\u5c16\u7814\u7a76\u5458\u548c\u5b9e\u9a8c\u5ba4\u65e9\u5df2\u77e5\u9053 Surge \u8d28\u91cf\u6700\u9ad8\u3001\u89c4\u6a21\u6700\u5927\u3001\u901f\u5ea6\u6700\u5feb\uff0c\u4f46\u6574\u4e2a AI \u9886\u57df\u6269\u5f20\u592a\u5feb\u3001\u6bcf\u5929\u6d8c\u5165\u65b0\u4eba\uff0cScale \u6536\u8d2d\u628a\u805a\u5149\u706f\u6253\u5230\u4e86\u4ed6\u4eec\u8eab\u4e0a\uff0c\u4e00\u591c\u4e4b\u95f4\u5927\u91cf\u65b0\u56e2\u961f\u5e26\u6765\u65b0\u9700\u6c42\u3002\u6587\u5316\u4e0a\uff0c\u8fd9\u79cd\u5f71\u54cd\u529b\u6700\u96c6\u4e2d\u5730\u4f53\u73b0\u5728\u62db\u8058\uff1aSurge \u66f4\u50cf\u4e00\u4e2a\u7814\u7a76\u5b9e\u9a8c\u5ba4\uff0c\u800c\u975e\u8ffd\u9010\u91d1\u94b1\u3001\u7092\u4f5c\u548c\u4f30\u503c\u7684\u666e\u901a\u7845\u8c37\u521b\u4e1a\u516c\u53f8\uff0c\u4ed6\u4eec\u60f3\u62db\u7684\u662f&#8221;\u9aa8\u5b50\u91cc\u7684\u7814\u7a76\u8005&#8221;\uff0c\u771f\u6b63\u5173\u5fc3\u6570\u636e\u3001AI \u548c\u6a21\u578b\u8fdb\u6b65\uff0c\u800c\u4e0d\u662f\u589e\u957f\u9ed1\u5ba2\u5f0f\u3001\u4e3a\u63d0\u5347\u6536\u5165\u4e0d\u60dc\u63a8\u9500\u4e0d\u9700\u8981\u751a\u81f3\u65e0\u7528\u4ea7\u54c1\u7684\u4eba\u3002\u540e\u8005\u5f80\u5f80\u50cf\u63a8\u9500\u5458\uff0c\u800c\u4e0d\u662f\u6df1\u5165\u6a21\u578b\u95ee\u9898\u3001\u5e2e\u5ba2\u6237\u6b63\u786e\u5ea6\u91cf\u6a21\u578b\u8fdb\u6b65\u7684\u5408\u4f5c\u8005\u3002\u8fd9\u79cd&#8221;\u628a\u6a21\u578b\u8fdb\u6b65\u7f6e\u4e8e\u6536\u5165\u4e4b\u4e0a&#8221;\u7684\u4fe1\u5ff5\uff0c\u6df1\u523b\u5851\u9020\u4e86\u516c\u53f8\u3002<\/p>\n<h2>\u5341\u4e8c\u3001\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u7684\u5206\u6b67\uff1a\u662f\u5426\u7406\u776c Arena\uff0c\u4ee5\u53ca\u4f18\u5316\u76ee\u6807\u4e4b\u4e89<\/h2>\n<p>\u5728\u6a21\u578b\u8fdb\u6b65\u8def\u5f84\u4e0a\uff0cEdwin \u770b\u5230\u7684\u5206\u6b67\u6bd4\u9884\u60f3\u5927\u5f97\u591a\uff0c\u51e0\u4e4e\u6bcf\u4e2a\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u90fd\u6709\u81ea\u5df1\u7684\u4e00\u5957\u3002\u9ad8\u5c42\u4e0a\u5206\u6b67\u96c6\u4e2d\u5728\u4e24\u5904\uff1a\u9009\u5b8c\u76ee\u6807\u51fd\u6570\u540e\uff0c\u7528\u4ec0\u4e48\u8bad\u7ec3\u7b97\u6cd5\u3001\u91c7\u4ec0\u4e48\u6570\u636e\uff1b\u4ee5\u53ca\u6700\u88ab\u4f4e\u4f30\u7684\u2014\u2014\u5404\u81ea\u4f18\u5316\u4ec0\u4e48\u3001\u5173\u6ce8\u4ec0\u4e48\u3002<\/p>\n<p>\u5c31 LLM Arena \u800c\u8a00\uff0c\u4e00\u4e9b\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u5e72\u8106\u9009\u62e9\u5b8c\u5168\u4e0d\u7406\u4f1a\u5b83\uff0c\u800c\u8fd9\u4e9b\u5b9e\u9a8c\u5ba4\u53cd\u800c\u505a\u5f97\u66f4\u597d\u3002\u56e0\u4e3a\u4e0d\u5f97\u4e0d\u7406\u776c Arena \u7684\u5b9e\u9a8c\u5ba4\uff0c\u5176\u7814\u7a76\u5458\u6e05\u695a\u4f18\u5316 Arena \u4f1a\u5e26\u6765\u8d1f\u5411\u8fdb\u6b65\u3001\u4f1a\u50ac\u751f\u5e7b\u89c9\u2014\u2014Arena \u7528\u6237\u5e76\u4e0d\u5728\u4e4e\u5e7b\u89c9\uff0c\u5e7b\u89c9\u53cd\u800c\u8ba9\u56de\u7b54\u663e\u5f97\u72c2\u91ce\u6709\u8da3\u3001\u50cf\u5c0f\u62a5\u4e00\u6837\u8bf1\u4eba\u3002\u90a3\u4e9b\u6709\u575a\u5b9a AGI \u8def\u5f84\u4fe1\u5ff5\u3001\u4e0d\u5fc5\u8ffd\u9010\u516c\u5f00\u7092\u4f5c\u6392\u884c\u699c\u7684\u5b9e\u9a8c\u5ba4\uff0c\u53cd\u800c\u83b7\u5f97\u4e86\u81ea\u7531\uff0c\u8fd9\u5728\u5f88\u5927\u7a0b\u5ea6\u4e0a\u5851\u9020\u4e86\u6a21\u578b\u8fdb\u6b65\u3002\u66f4\u6e05\u6670\u7684\u5206\u6b67\u4f53\u73b0\u5728\u4f18\u5316\u76ee\u6807\u4e0a\uff1aOpenAI \u5982\u4eca\u66f4\u504f\u5411\u4f18\u5316\u7528\u6237\u53c2\u4e0e\u5ea6\u2014\u2014\u8d85\u957f\u4f1a\u8bdd\u3001\u65e5\u6d3b\u7528\u6237\uff1b\u800c Anthropic \u66f4\u504f\u5411\u4f18\u5316\u751f\u4ea7\u529b\u2014\u2014\u901a\u8fc7\u4e0e\u6a21\u578b\u4ea4\u4e92\u80fd\u521b\u9020\u591a\u5c11\u4ef7\u503c\u3001\u8282\u7701\u591a\u5c11\u65f6\u95f4\u3001\u8fd1\u4e4e GDP \u7684\u4ea7\u51fa\u3002\u76ee\u6807\u4e0d\u540c\uff0c\u5851\u9020\u4e86\u4ea7\u54c1\u7c7b\u578b\u3001\u5438\u5f15\u7684\u4eba\u624d\u548c\u6a21\u578b\u80fd\u529b\u3002<\/p>\n<h2>\u5341\u4e09\u3001\u4ece&#8221;\u4e00\u7edf\u6a21\u578b&#8221;\u5230&#8221;\u6a21\u578b\u661f\u5ea7&#8221;\uff0c\u6bcf\u5bb6\u516c\u53f8\u90fd\u8be5\u6709\u81ea\u5df1\u7684\u8bba\u70b9<\/h2>\n<p>Edwin \u5766\u8a00\u81ea\u5df1\u8fc7\u53bb\u76f8\u4fe1&#8221;\u4e00\u4e2a\u4e00\u7edf\u5929\u4e0b\u7684\u6a21\u578b&#8221;\uff08one model to rule them all\uff09\u2014\u2014\u4e00\u4e2a\u8d85\u7ea7\u667a\u80fd\u6a21\u578b\u80fd\u65e0\u7f1d\u5207\u6362\u3001\u9002\u5e94\u4f60\u60f3\u505a\u7684\u4e00\u5207\u3002\u4f46\u8fc7\u53bb\u4e00\u5e74\u4ed6\u610f\u8bc6\u5230\uff0c\u4e16\u754c\u5982\u6b64\u4e30\u5bcc\uff0c\u6c38\u8fdc\u4e0d\u4f1a\u6709\u4e00\u52b3\u6c38\u9038\u7684\u65b9\u6848\uff1b\u6bcf\u5bb6\u516c\u53f8\u3001\u6bcf\u4e2a\u5b9e\u9a8c\u5ba4\u3001\u6bcf\u4e2a AI \u90fd\u9700\u8981\u6709\u81ea\u5df1\u7684\u8bba\u70b9\uff08thesis\uff09\uff1a\u4ec0\u4e48\u6837\u7684 AI \u5bf9\u4e16\u754c\u6709\u7528\u3001\u600e\u6837\u6700\u597d\u5730\u670d\u52a1\u4e8e\u4eba\uff0c\u800c\u8fd9\u4e2a\u8bba\u70b9\u4f1a\u5851\u9020\u6a21\u578b\u7684\u884c\u4e3a\u3002\u4e24\u4e2a\u6a21\u578b\u53ef\u4ee5\u540c\u6837\u806a\u660e\uff0c\u5374\u6709\u4e0d\u540c\u4eba\u683c\u3001\u4e0d\u540c\u56de\u7b54\u504f\u89c1\u3001\u4e0d\u540c\u4ea4\u8c08\u65b9\u5f0f\uff1b\u6b63\u5982 Google \u505a\u793e\u4ea4\u4e0e Facebook \u505a\u641c\u7d22\u5fc5\u7136\u622a\u7136\u4e0d\u540c\uff0cAI \u4e5f\u4f1a\u5982\u6b64\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/frontier-lab-objectives-diverge.jpg\" alt=\"Frontier labs divergent objectives constellation\"\/><figcaption>Model constellation era<\/figcaption><\/figure>\n<p>\u56e0\u6b64\u4ed6\u8ba4\u4e3a\u5e94\u8be5\u6709\u66f4\u591a\u4eba\u5efa\u6a21\u578b\uff0c\u56e0\u4e3a\u4e16\u754c\u9700\u8981\u591a\u79cd\u8bba\u70b9\uff1b\u5927\u578b\u91d1\u878d\u3001\u533b\u7597\u673a\u6784\u6700\u7ec8\u90fd\u5e94\u8bad\u7ec3\u81ea\u5df1\u7684\u6a21\u578b\uff0c\u56e0\u4e3a\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u4f18\u5316\u7684\u672a\u5fc5\u662f\u4f60\u4f18\u5316\u7684\uff0c\u800c\u6a21\u578b\u5982\u6b64\u91cd\u8981\u3001\u8981\u90e8\u7f72\u5230 99.9999% \u7684\u7528\u4f8b\uff0c\u5c31\u8be5\u8ffd\u6c42\u6700\u597d\u7684\u4ef7\u503c\u4e0e\u8868\u73b0\u3002\u4eca\u5929\u8d70\u5230 SOTA \u4ecd\u76f8\u5f53\u6602\u8d35\uff0c\u4f46\u4e00\u4e2a\u6709\u81ea\u5df1\u4ea7\u54c1\u8bba\u70b9\u7684&#8221;\u6709\u4e3b\u89c1\u7684\u6a21\u578b&#8221;\uff0c\u5373\u4fbf\u843d\u540e SOTA \u516d\u5230\u5341\u4e8c\u4e2a\u6708\u4e5f\u5b8c\u5168\u53ef\u884c\u3002\u4ed6\u4e5f\u786e\u8ba4\uff1a\u6a21\u578b\u4ecd\u5728\u7a33\u6b65\u53d8\u597d\uff0c\u53ea\u662f\u6700\u8fd1\u4e09\u516d\u4e2a\u6708\uff0c\u8fdb\u6b65\u4e0d\u518d\u50cf\u65e9\u671f\u90a3\u6837\u5bf9\u7528\u6237\u663e\u800c\u6613\u89c1\uff1b\u4ed6\u81ea\u5df1\u6700\u8fd1\u5927\u91cf\u7528 Claude \u5199\u4f5c\uff0c\u76f8\u6bd4\u51e0\u4e2a\u6708\u524d\u7684\u63d0\u5347\u4ee4\u4ed6\u9707\u60ca\u3002<\/p>\n<h2>\u5341\u56db\u3001\u591a\u6a21\u6001\uff1a\u89c6\u9891\u3001\u673a\u5668\u4eba\u3001\u751f\u7269\uff0c50% \u4ee5\u4e0a\u5de5\u4f5c\u5df2\u5728\u7eaf\u6587\u672c\u4e4b\u5916<\/h2>\n<p>\u88ab\u95ee\u53ca\u89c6\u9891\u3001\u673a\u5668\u4eba\u3001\u751f\u7269\u7b49\u591a\u6a21\u6001\u65b9\u5411\uff0cEdwin \u900f\u9732 Surge \u4eca\u5929\u8d85\u8fc7 50% \u7684\u5de5\u4f5c\u5df2\u7ecf\u5728\u7eaf\u6587\u672c\u4e4b\u5916\uff0c\u8fd9\u4e0e&#8221;\u4e0d\u60dc\u4e00\u5207\u4ee3\u4ef7\u4f7f\u80fd AGI&#8221;\u7684\u4f7f\u547d\u4e00\u81f4\u2014\u2014\u8981\u8ba9 AI \u5728\u73b0\u5b9e\u4e16\u754c\u6709\u7528\uff0c\u5c31\u5fc5\u987b\u7406\u89e3\u5e76\u8de8\u8fd9\u4e9b\u9886\u57df\u64cd\u4f5c\u3002\u4eba\u4eec\u5f80\u5f80\u4f4e\u4f30\u8d28\u91cf\u5728\u8fd9\u4e9b\u770b\u4f3c\u7b80\u5355\u6a21\u6001\u4e2d\u7684\u5206\u91cf\uff1a\u5373\u4f7f\u662f\u8bbe\u8ba1 prompt\uff0c\u4e5f\u9700\u8981\u5927\u91cf\u521b\u9020\u529b\u4e0e\u6280\u672f\u53bb\u63a2\u7d22\u6574\u4e2a\u80fd\u529b\u5206\u5e03\u3002\u8d28\u91cf\u5fc5\u987b\u8d85\u8d8a\u673a\u68b0\u7684\u6307\u4ee4\u9075\u5faa\u548c\u673a\u68b0\u7684\u6b63\u786e\u6027\u3002<\/p>\n<p>\u4ed6\u7528 Scorsese \u4e0e\u9ad8\u4e2d\u827a\u672f\u6bd5\u4e1a\u751f\u90fd\u62cd&#8221;\u4e00\u6761\u9c7c&#8221;\u7684\u7535\u5f71\u4f5c\u6bd4\uff1a\u4e24\u4eba\u90fd\u80fd\u5b8c\u6210\uff0c\u4f46 Scorsese \u7684\u9c7c\u7535\u5f71\u663e\u7136\u66f4\u597d\u2014\u2014\u8fd9\u5c31\u662f\u54c1\u5473\u3001\u6210\u719f\u5ea6\u3001\u521b\u9020\u529b\u4e0e&#8221;\u66f4\u8fdb\u4e00\u6b65&#8221;\u7684\u5dee\u522b\u3002\u7528\u6237\u60f3\u8981\u7684\u4e0d\u662f\u5b57\u9762\u6267\u884c\u6307\u4ee4\uff0c\u800c\u662f\u4e00\u4ef6\u4ee4\u4eba\u60ca\u8273\u3001\u5bcc\u6709\u60f3\u8c61\u4e0e\u521b\u9020\u529b\u3001\u62c9\u9ad8\u6807\u51c6\u7684\u4f5c\u54c1\u3002\u5bf9\u4e8e\u9700\u8981\u786c\u4ef6\u91c7\u96c6\u6570\u636e\u7684\u673a\u5668\u4eba\u3001\u751f\u7269\u7b49\u7a7a\u95f4\uff0cSurge \u7684\u601d\u8def\u662f\uff1a\u4e0d\u60dc\u4e00\u5207\u4ee3\u4ef7\u4f7f\u80fd\u52a0\u901f AGI \u6240\u9700\u7684\u6570\u636e\uff0c\u6709\u65f6\u8981\u9020\u65b0\u5de5\u5177\u3001\u4e70\u65b0\u786c\u4ef6\u3001\u6269\u5c55\u5230\u4efb\u4f55\u9700\u8981\u7684\u7a7a\u95f4\uff1b\u4ed6\u4eec\u662f\u6280\u672f\u516c\u53f8\uff0c\u4e0d\u4f1a\u56e0\u4e3a\u81ea\u5df1\u53ea\u662f&#8221;pivot \u8fdb\u6765&#8221;\u5c31\u628a\u81ea\u5df1\u7a84\u5316\uff0c\u53cd\u800c\u6bd4\u5f88\u591a\u540c\u884c\u66f4\u957f\u671f\u5730\u601d\u8003\u6240\u9700\u7684\u4e00\u5207\u3002<\/p>\n<h2>\u5341\u4e94\u3001\u5feb\u95ee\u5feb\u7b54\uff1a\u6700\u5927\u89c2\u5ff5\u8f6c\u53d8\u3001\u6700\u5927\u9057\u61be\u3001\u7406\u60f3\u4f18\u5316\u76ee\u6807<\/h2>\n<p>\u4e00\u5e74\u6765\u5bf9 AI \u6700\u5927\u7684\u89c2\u5ff5\u8f6c\u53d8\uff1a\u4ed6\u8fc7\u53bb\u76f8\u4fe1&#8221;\u4e00\u7edf\u6a21\u578b&#8221;\uff0c\u5982\u4eca\u771f\u6b63\u7406\u89e3\u4e0d\u540c\u7684\u4ea7\u54c1\u89c2\u70b9\u3001AI \u89c2\u70b9\u5c06\u5851\u9020\u672a\u6765\u3002\u5efa\u7acb\u516c\u53f8\u6700\u5927\u7684\u9057\u61be\uff1a\u4ed6\u51fa\u8eab\u7814\u7a76\u4e0e\u6570\u636e\u79d1\u5b66\uff0c\u539f\u672c\u70ed\u7231\u5199\u4f5c\u3001\u53d1\u535a\u5ba2\u3001\u5206\u4eab\u6d1e\u89c1\uff0c\u65e9\u671f\u505a\u5f97\u5f88\u591a\uff0c\u540e\u6765\u5fd9\u5230\u505c\u6389\uff0c\u8fd1\u4e24\u4e09\u5e74\u516c\u5f00\u5206\u4eab\u660e\u663e\u51cf\u5c11\uff0c\u8fd9\u662f\u4ed6\u6700\u5927\u7684\u9519\u8bef\uff0c\u73b0\u5728\u5e0c\u671b\u91cd\u65b0\u6062\u590d\u3002\u5982\u679c\u6709\u4e00\u5468\u5047\u671f\u5199\u4e00\u7bc7\u957f\u6587\uff0c\u4ed6\u6700\u60f3\u5199\u7684\u4e3b\u9898\u662f&#8221;\u76ee\u6807\u51fd\u6570&#8221;\u2014\u2014\u6bcf\u4e2a\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u7a76\u7adf\u5728\u4f18\u5316\u4ec0\u4e48\uff0c\u8fd9\u4e2a\u6982\u5ff5\u5fae\u5999\u5374\u5f71\u54cd\u6df1\u8fdc\uff1a\u53c2\u4e0e\u5ea6\uff1f\u6709\u7528\u6027\uff1f\u7528\u6237\u6570\uff1f\u8fd8\u662f GDP\uff1f\u82e5\u8ba9\u4ed6\u638c\u7ba1\u4e00\u4e2a\u5b9e\u9a8c\u5ba4\uff0c\u4ed6\u4f1a\u4f18\u5316&#8221;\u4e00\u4e2a\u6708\u540e\u4f60\u662f\u5426\u4f1a\u56e0\u4e3a\u8fd9\u6b21\u4e0e\u6a21\u578b\u7684\u4ea4\u4e92\u800c\u9ad8\u5174\u3001\u5b83\u662f\u5426\u4ee5\u67d0\u79cd\u65b9\u5f0f\u6539\u53d8\u4e86\u4f60\u7684\u4eba\u751f&#8221;\u2014\u2014\u6bd4\u5982\u5728\u5047\u671f\u5411\u4f60\u63a8\u8350\u4e86\u4e00\u4e2a\u4ece\u6ca1\u60f3\u8fc7\u7684\u5730\u70b9\uff0c\u6216\u5728\u533b\u7597\u95ee\u9898\u4e0a\u654f\u9510\u6ce8\u610f\u5230\u4f60\u672c\u6ca1\u8bf4\u6e05\u7684\u9690\u60a3\u3002\u4ed6\u6700\u540e\u6307\u51fa\uff0c\u8fd9\u4e9b\u95ee\u9898\u4e0e\u793e\u4f1a\u957f\u671f\u9762\u4e34\u7684\u96be\u9898\u540c\u6784\uff0cAI \u9762\u4e34\u7684\u4f18\u5316\u98ce\u9669\u4e0e\u793e\u4ea4\u5a92\u4f53\u9ad8\u5ea6\u5e73\u884c\uff0c\u800c\u968f\u7740\u6a21\u578b\u8fdb\u6b65\uff0c\u8fd9\u4e2a\u95ee\u9898\u53ea\u4f1a\u6108\u53d1\u91cd\u8981\u3002<\/p>\n<p>\u5bf9\u8bdd\u7ed3\u5c3e\uff0cEdwin \u63a8\u8350\u5927\u5bb6\u5173\u6ce8 Surge \u7684\u535a\u5ba2\uff0c\u516c\u53f8\u6b63\u6062\u590d\u66f4\u9891\u7e41\u7684\u5199\u4f5c\uff0c\u5206\u4eab\u66f4\u591a\u6d1e\u89c1\u4e0e\u5206\u6790\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u672c\u671f\u300aUnsupervised Learning\u300b\u7531 Jacob Effron \u4e3b\u6301\uff0c\u5bf9\u8bdd\u5609\u5bbe\u662f\u4f30\u503c\u7ea6 240 &#8230; <a title=\"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3\" class=\"read-more\" href=\"https:\/\/aicrunchx.com\/cn\/archives\/67\" aria-label=\"\u9605\u8bfb Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3\">\u9605\u8bfb\u66f4\u591a<\/a><\/p>\n","protected":false},"author":1,"featured_media":60,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[122,116,107,110,111,26,133,115,113,128,33,132,120,117,119,118,121,129,127,131,100,114,124,109,134,123,130,108,112,126,125,135],"class_list":["post-67","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-inference","tag-ai-slop","tag-benchmark-overfitting","tag-edwin-chen","tag-facebook","tag-gaia","tag-google","tag-hemingway","tag-human-evaluation","tag-mcp","tag-meritocracy","tag-meta","tag-mit","tag-model-constellation","tag-model-trajectories","tag-multimodal","tag-objective-function","tag-pivot-culture","tag-product-thesis","tag-productivity","tag-pytorch","tag-reward-hacking","tag-rl-environments","tag-rlhf","tag-scale-ai","tag-scorsese","tag-sft","tag-street-smarts","tag-surge","tag-swe-bench","tag-user-engagement","tag-verifier","tag-yc"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/aicrunchx.com\/cn\/archives\/67\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX\" \/>\n<meta property=\"og:description\" content=\"\u672c\u671f\u300aUnsupervised Learning\u300b\u7531 Jacob Effron \u4e3b\u6301\uff0c\u5bf9\u8bdd\u5609\u5bbe\u662f\u4f30\u503c\u7ea6 240 ... \u9605\u8bfb\u66f4\u591a\" \/>\n<meta property=\"og:url\" content=\"https:\/\/aicrunchx.com\/cn\/archives\/67\" \/>\n<meta property=\"og:site_name\" content=\"AICrunchX\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-21T02:46:27+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-10-21T04:32:02+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"640\" \/>\n\t<meta property=\"og:image:height\" content=\"360\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Gavin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"Gavin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67\"},\"author\":{\"name\":\"Gavin\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#\\\/schema\\\/person\\\/8e24b2f75248e5b0b64c19e28499a29e\"},\"headline\":\"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3\",\"datePublished\":\"2026-10-21T02:46:27+00:00\",\"dateModified\":\"2026-10-21T04:32:02+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67\"},\"wordCount\":201,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/sv-pivot-culture-criticism.jpg\",\"keywords\":[\"AI Slop\",\"Benchmark Overfitting\",\"Edwin Chen\",\"Facebook\",\"Gaia\",\"Google\",\"Hemingway\",\"Human Evaluation\",\"MCP\",\"Meritocracy\",\"Meta\",\"MIT\",\"Model Constellation\",\"Model Trajectories\",\"Multimodal\",\"Objective Function\",\"Pivot Culture\",\"Product Thesis\",\"Productivity\",\"PyTorch\",\"Reward Hacking\",\"RL Environments\",\"RLHF\",\"Scale AI\",\"Scorsese\",\"SFT\",\"Street Smarts\",\"Surge\",\"SWE-bench\",\"User Engagement\",\"Verifier\",\"YC\"],\"articleSection\":[\"AI Inference\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67\",\"url\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67\",\"name\":\"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/sv-pivot-culture-criticism.jpg\",\"datePublished\":\"2026-10-21T02:46:27+00:00\",\"dateModified\":\"2026-10-21T04:32:02+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#primaryimage\",\"url\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/sv-pivot-culture-criticism.jpg\",\"contentUrl\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/sv-pivot-culture-criticism.jpg\",\"width\":640,\"height\":360},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/67#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#website\",\"url\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/\",\"name\":\"AICrunchX\",\"description\":\"AI\u4e16\u754c\",\"publisher\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#organization\",\"name\":\"AICrunchX\",\"url\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"\",\"contentUrl\":\"\",\"caption\":\"AICrunchX\"},\"image\":{\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#\\\/schema\\\/logo\\\/image\\\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/#\\\/schema\\\/person\\\/8e24b2f75248e5b0b64c19e28499a29e\",\"name\":\"Gavin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g\",\"caption\":\"Gavin\"},\"sameAs\":[\"https:\\\/\\\/aicrunchx.com\\\/cn\"],\"url\":\"https:\\\/\\\/aicrunchx.com\\\/cn\\\/archives\\\/author\\\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/aicrunchx.com\/cn\/archives\/67","og_locale":"zh_CN","og_type":"article","og_title":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX","og_description":"\u672c\u671f\u300aUnsupervised Learning\u300b\u7531 Jacob Effron \u4e3b\u6301\uff0c\u5bf9\u8bdd\u5609\u5bbe\u662f\u4f30\u503c\u7ea6 240 ... \u9605\u8bfb\u66f4\u591a","og_url":"https:\/\/aicrunchx.com\/cn\/archives\/67","og_site_name":"AICrunchX","article_published_time":"2026-10-21T02:46:27+00:00","article_modified_time":"2026-10-21T04:32:02+00:00","og_image":[{"width":640,"height":360,"url":"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg","type":"image\/jpeg"}],"author":"Gavin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"Gavin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#article","isPartOf":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67"},"author":{"name":"Gavin","@id":"https:\/\/aicrunchx.com\/cn\/#\/schema\/person\/8e24b2f75248e5b0b64c19e28499a29e"},"headline":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3","datePublished":"2026-10-21T02:46:27+00:00","dateModified":"2026-10-21T04:32:02+00:00","mainEntityOfPage":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67"},"wordCount":201,"commentCount":0,"publisher":{"@id":"https:\/\/aicrunchx.com\/cn\/#organization"},"image":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#primaryimage"},"thumbnailUrl":"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg","keywords":["AI Slop","Benchmark Overfitting","Edwin Chen","Facebook","Gaia","Google","Hemingway","Human Evaluation","MCP","Meritocracy","Meta","MIT","Model Constellation","Model Trajectories","Multimodal","Objective Function","Pivot Culture","Product Thesis","Productivity","PyTorch","Reward Hacking","RL Environments","RLHF","Scale AI","Scorsese","SFT","Street Smarts","Surge","SWE-bench","User Engagement","Verifier","YC"],"articleSection":["AI Inference"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/aicrunchx.com\/cn\/archives\/67#respond"]}]},{"@type":"WebPage","@id":"https:\/\/aicrunchx.com\/cn\/archives\/67","url":"https:\/\/aicrunchx.com\/cn\/archives\/67","name":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3 - AICrunchX","isPartOf":{"@id":"https:\/\/aicrunchx.com\/cn\/#website"},"primaryImageOfPage":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#primaryimage"},"image":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#primaryimage"},"thumbnailUrl":"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg","datePublished":"2026-10-21T02:46:27+00:00","dateModified":"2026-10-21T04:32:02+00:00","breadcrumb":{"@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/aicrunchx.com\/cn\/archives\/67"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#primaryimage","url":"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg","contentUrl":"https:\/\/aicrunchx.com\/cn\/wp-content\/uploads\/2026\/10\/sv-pivot-culture-criticism.jpg","width":640,"height":360},{"@type":"BreadcrumbList","@id":"https:\/\/aicrunchx.com\/cn\/archives\/67#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/aicrunchx.com\/cn\/"},{"@type":"ListItem","position":2,"name":"Surge CEO Edwin Chen\u6df1\u5ea6\u5bf9\u8bdd\uff1a\u522b\u88abLLM Arena\u5e26\u504f\uff0c\u771f\u6b63\u7684\u4eba\u5de5\u8bc4\u4f30\u3001RL\u73af\u5883\u4e0e\u6a21\u578b\u661f\u5ea7\u65f6\u4ee3"}]},{"@type":"WebSite","@id":"https:\/\/aicrunchx.com\/cn\/#website","url":"https:\/\/aicrunchx.com\/cn\/","name":"AICrunchX","description":"AI\u4e16\u754c","publisher":{"@id":"https:\/\/aicrunchx.com\/cn\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/aicrunchx.com\/cn\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Organization","@id":"https:\/\/aicrunchx.com\/cn\/#organization","name":"AICrunchX","url":"https:\/\/aicrunchx.com\/cn\/","logo":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/aicrunchx.com\/cn\/#\/schema\/logo\/image\/","url":"","contentUrl":"","caption":"AICrunchX"},"image":{"@id":"https:\/\/aicrunchx.com\/cn\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/aicrunchx.com\/cn\/#\/schema\/person\/8e24b2f75248e5b0b64c19e28499a29e","name":"Gavin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/secure.gravatar.com\/avatar\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/d6683a4d754e1836ca51137c3144cfb0ce071c46d677c7510cfd45a5214c8c92?s=96&d=mm&r=g","caption":"Gavin"},"sameAs":["https:\/\/aicrunchx.com\/cn"],"url":"https:\/\/aicrunchx.com\/cn\/archives\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/posts\/67","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/comments?post=67"}],"version-history":[{"count":1,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/posts\/67\/revisions"}],"predecessor-version":[{"id":68,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/posts\/67\/revisions\/68"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/media\/60"}],"wp:attachment":[{"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/media?parent=67"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/categories?post=67"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aicrunchx.com\/cn\/wp-json\/wp\/v2\/tags?post=67"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}