{"id":1104,"date":"2026-11-04T20:23:59","date_gmt":"2026-11-04T12:23:59","guid":{"rendered":"https:\/\/www.timedigital.cn\/new\/1104\/"},"modified":"2026-11-04T20:23:59","modified_gmt":"2026-11-04T12:23:59","slug":"%e6%8b%9b%e5%95%86%e5%b1%80%e7%8b%ae%e5%ad%90%e5%b1%b1%e5%ae%9e%e9%aa%8c%e5%ae%a4%e5%8f%91%e5%b8%83%e3%80%8c%e9%a9%ad%e7%89%a9%e3%80%8d%e5%85%b7%e8%ba%ab%e6%8e%a8%e7%90%86%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"https:\/\/www.timedigital.cn\/new\/1104\/","title":{"rendered":"\u62db\u5546\u5c40\u72ee\u5b50\u5c71\u5b9e\u9a8c\u5ba4\u53d1\u5e03\u300c\u9a6d\u7269\u2122\u300d\u5177\u8eab\u63a8\u7406\u6a21\u578b"},"content":{"rendered":"<div class=\"tdx-art\">\n<p class=\"tdx-lead\">\u673a\u5668\u4eba\u5206\u4e0d\u6e05\u753b\u9762\u91cc&#8221;\u66f4\u9760\u4e0a&#8221;\u548c\u771f\u5b9e\u4e09\u7ef4\u7a7a\u95f4&#8221;\u66f4\u9ad8&#8221;\u7684\u5dee\u522b\uff0c\u89c6\u89d2\u4e00\u6362\u5c31\u4f1a\u8bef\u5224\u7269\u4f53\u8eab\u4efd\u3001\u76f8\u673a\u8fd0\u52a8\u6216\u4efb\u52a1\u8fdb\u5ea6\u3002\u62db\u5546\u5c40\u5148\u8fdb\u6280\u672f\u7814\u7a76\u9662\u4e0b\u5c5e\u72ee\u5b50\u5c71\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u8fd1\u65e5\u53d1\u5e03\u5177\u8eab\u63a8\u7406\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\u300c\u9a6d\u7269\u2122 Lion-ER-35B-A3B\u300d\uff0c<strong>\u603b\u53c2\u6570\u7ea635B\uff0c\u6bcf\u4e2aToken\u4ec5\u6fc0\u6d3b\u7ea63B\u53c2\u6570\uff0c\u5728\u5341\u9879\u4e3b\u6d41\u5177\u8eab\u76f8\u5173\u57fa\u51c6\u4e0a\u5168\u90e8\u6307\u6807\u4e0a\u6da8<\/strong>\uff0c\u7b49\u6743\u7efc\u5408\u5747\u5206\u8fbe\u523065.29\u3002<\/p>\n<p>\u8be5\u6a21\u578b\u8d70\u7684\u662f\u80fd\u529b\u7275\u5f15\u7684\u6570\u636e\u6280\u672f\u8def\u7ebf\uff0c\u4e0d\u9760\u76f2\u76ee\u5806\u8bad\u7ec3\u6570\u636e\u3002\u56e2\u961f\u628a\u5177\u8eab\u63a8\u7406\u62c6\u89e3\u4e3a\u5b9e\u4f53\u3001\u7a7a\u95f4\u3001\u65f6\u5e8f\u3001\u7269\u7406\u3001\u53ef\u4f9b\u6027\u3001\u4efb\u52a1\u3001\u793e\u4f1a\u3001\u8bc1\u636e8\u4e2a\u4e00\u7ea7\u9886\u57df\uff0c\u7ec6\u5316\u51fa39\u9879\u4e8c\u7ea7\u80fd\u529b\uff0c\u518d\u628a\u63a8\u7406\u65b9\u5f0f\u3001\u8f93\u5165\u5f62\u5f0f\u548c\u56de\u7b54\u5f62\u5f0f\u4f5c\u4e3a\u72ec\u7acb\u7ef4\u5ea6\u5206\u5f00\u6807\u6ce8\u3002\u80fd\u529b\u84dd\u56fe\u9884\u5148\u89c4\u5212\u9700\u8981\u8986\u76d6\u7684\u80fd\u529b\u8303\u56f4\uff0c\u8bc4\u6d4b\u627f\u62c5\u67e5\u6f0f\u8865\u7f3a\u7684\u89d2\u8272\u3002<\/p>\n<h2>\u80fd\u529b\u7f16\u8bd1<\/h2>\n<p>\u80fd\u529b\u7f16\u8bd1\u628a\u62bd\u8c61\u80fd\u529b\u76ee\u6807\u9010\u9879\u843d\u5730\u4e3a\u573a\u666f\u3001\u89c2\u5bdf\u65b9\u5f0f\u3001\u63d0\u95ee\u548c\u7b54\u6848\u4f9d\u636e\u3002\u8de8\u89c6\u89d2\u63a8\u7406\u8981\u5148\u660e\u786e\u89c6\u89d2\u6570\u91cf\u3001\u9700\u8981\u5728\u591a\u5f20\u753b\u9762\u95f4\u5efa\u7acb\u5173\u8054\u7684\u76ee\u6807\u7269\u4f53\u3001\u5224\u5b9a\u65b9\u4f4d\u6240\u91c7\u7528\u7684\u89c2\u6d4b\u8005\u4f4d\u7f6e\u4e0e\u671d\u5411\uff0c\u540c\u65f6\u6838\u9a8c\u89c6\u89c9\u7d20\u6750\u662f\u5426\u8db3\u4ee5\u652f\u6491\u6a21\u578b\u63a8\u5bfc\u51fa\u7b54\u6848\u3002\u5173\u952e\u89c6\u89d2\u7f3a\u5931\u3001\u8bc1\u636e\u4e0d\u8db3\u65f6\uff0c\u8fd9\u9053\u9898\u53ea\u80fd\u4f9d\u9760\u731c\u6d4b\uff0c\u4e0d\u80fd\u7eb3\u5165\u5408\u683c\u8bad\u7ec3\u6837\u672c\u3002<\/p>\n<p>\u8bad\u7ec3\u9898\u76ee\u6709\u4e09\u6761\u6784\u9898\u8def\u5f84\uff1a\u4ece\u5df2\u6709QA\u4e0e\u89c6\u89c9\u6807\u6ce8\u4e2d\u8f6c\u6362\u3001\u4ece\u89c2\u5bdf\u7d20\u6750\u4e2d\u53d1\u73b0\u8bc1\u636e\u3001\u4ece\u7ed3\u6784\u5316\u6761\u4ef6\u4e0e\u5173\u7cfb\u4e2d\u63a8\u5bfc\u3002\u76d1\u7763\u6838\u9a8c\u91c7\u7528\u6570\u636e\u63d0\u7eaf\u3001\u51e0\u4f55\u6784\u9898\u3001\u5b9a\u5411\u5408\u6210\u4e09\u6761\u4e92\u8865\u8def\u7ebf\uff0c\u5176\u4e2d\u5b9a\u5411\u5408\u6210\u751f\u6210\u7684\u95ee\u9898\u4e0e\u6807\u6ce8\u5fc5\u987b\u7ecf\u8fc7\u72ec\u7acb\u4f5c\u7b54\u6216\u989d\u5916\u5ba1\u6838\uff0c\u751f\u6210\u6a21\u578b\u7ed9\u51fa\u7684\u7b54\u6848\u4e0d\u76f4\u63a5\u4f5c\u4e3a\u771f\u503c\u3002\u6a21\u578b\u7b54\u9519\u4e0d\u662f\u5220\u6837\u672c\u7684\u7406\u7531\uff0c\u591a\u4e2a\u6a21\u578b\u7ed9\u51fa\u76f8\u540c\u7b54\u6848\u4e5f\u4e0d\u80fd\u5355\u72ec\u8bc1\u660e\u7b54\u6848\u6b63\u786e\u3002\u5bf9\u6301\u7eed\u5931\u8bef\u7684\u9ad8\u96be\u6837\u672c\uff0c\u56e2\u961f\u5f15\u5165Human in the loop\u4eba\u5de5\u590d\u6838\uff0c\u642d\u5efa\u4e86\u9762\u5411\u590d\u6742\u5177\u8eab\u4efb\u52a1\u7684\u53ef\u89c6\u5316\u6838\u9a8c\u5e73\u53f0\u3002<\/p>\n<h2>Eval-Driven Data Harness<\/h2>\n<p>\u76d1\u7763\u8bad\u7ec3\u96c6\u5171\u5305\u542b<strong>312380\u6761\u6837\u672c<\/strong>\uff0c\u8bad\u7ec3\u65f6\u6309\u80fd\u529b\u9700\u6c42\u5206\u914d\u9884\u7b97\uff0c\u800c\u4e0d\u662f\u7531\u6570\u636e\u6e90\u9898\u91cf\u51b3\u5b9a\u6743\u91cd\u3002Eval-Driven Data Harness\u628a\u8bc4\u6d4b\u5931\u8d25\u8f6c\u6210\u4e0b\u4e00\u8f6e\u6570\u636e\u9700\u6c42\uff1a\u5148\u6838\u67e5\u6807\u6ce8\u3001\u753b\u9762\u8bc1\u636e\u3001\u8bc4\u5206\u662f\u5426\u5b58\u5728\u95ee\u9898\uff0c\u518d\u533a\u5206\u5076\u53d1\u5931\u8bef\u4e0e\u7cfb\u7edf\u6027\u80fd\u529b\u77ed\u677f\uff0c\u786e\u8ba4\u5b58\u5728\u80fd\u529b\u7f3a\u53e3\u540e\u4ece\u8986\u76d6<strong>31\u4e2a\u6570\u636e\u6e90\u3001\u7ea6430\u4e07\u6761\u8bb0\u5f55<\/strong>\u7684\u7d20\u6750\u6c60\u4e2d\u68c0\u7d22\u63a8\u7406\u903b\u8f91\u76f8\u540c\u3001\u573a\u666f\u4e0d\u540c\u7684\u65b0\u7d20\u6750\uff0c\u7ecf\u6784\u9898\u6838\u9a8c\u540e\u52a0\u5165\u8bad\u7ec3\u3002\u4e00\u6b21\u8bc4\u6d4b\u5931\u8d25\uff0c\u7531\u6b64\u53d8\u6210\u4e00\u6279\u65b0\u7684\u8bad\u7ec3\u6570\u636e\u3002<\/p>\n<p>\u5341\u9879\u516c\u5f00\u57fa\u51c6\u91c7\u7528avg@4\u8bc4\u6d4b\u53e3\u5f84\uff0c\u6bcf\u9053\u9898\u72ec\u7acb\u4f5c\u7b544\u6b21\u53d6\u5e73\u5747\u5206\u3002Where2Place\u7a7a\u95f4\u653e\u7f6e\u8bc4\u6d4b\u5f97\u5206\u4e3a64.28\uff0c\u8f83\u57fa\u5ea7\u63d0\u53479.30\u5206\uff1bERQA\u5177\u8eab\u63a8\u7406\u57fa\u51c6\u5f97\u520664.38\uff0c\u63d0\u53476.19\u5206\uff1b\u7a7a\u95f4\u63a8\u7406\u57fa\u51c6SAT\u8fbe\u523089.33\uff1bERQA\u8f68\u8ff9\u7406\u89e3\u5b50\u7c7b68.18\uff0c\u63d0\u534712.12\u5206\u3002\u5728\u76f8\u673a\u8fd0\u52a8\u5224\u65ad\u7684\u516c\u5f00\u57fa\u51c6\u9898\u4e2d\uff0c\u57fa\u5ea7\u4ec5\u7b54\u5bf9\u4e00\u6b21\uff0c\u300c\u9a6d\u7269\u2122 Lion-ER-35B-A3B\u300d\u901a\u8fc7\u4e24\u5e27\u753b\u9762\u53d8\u5316\u89c4\u5f8b\u6b63\u786e\u5224\u65ad\u76f8\u673a\u8f6c\u52a8\u65b9\u5411\uff0c4\u6b21\u4f5c\u7b54\u5168\u90e8\u6b63\u786e\u3002<\/p>\n<p>\u76ee\u524d\u300c\u9a6d\u7269\u2122\u300d\u5df2\u5728Hugging Face\u5f00\u6e90\u6743\u91cd\uff0c\u91c7\u7528BF16 Safetensors\u683c\u5f0f\uff0c\u539f\u751f\u517c\u5bb9Transformers\u3001vLLM\u3001SGLang\u7b49\u4e3b\u6d41\u63a8\u7406\u6846\u67b6\u3002\u62db\u5546\u5c40\u72ee\u5b50\u5c71\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u8868\u793a\uff0c\u540e\u7eed\u5c06\u9762\u5411\u5177\u8eabAgent\u62d3\u5c55\u5de5\u5177\u8c03\u7528\u3001\u591a\u8f6e\u4ea4\u4e92\u4e0e\u6267\u884c\u53cd\u9988\u65b9\u9762\u7684\u8bad\u7ec3\u6570\u636e\uff0c\u63a2\u7d22\u957f\u7a0b\u4efb\u52a1\u4e2d\u7684\u8fde\u7eed\u51b3\u7b56\u4e0e\u5931\u8d25\u6062\u590d\uff0c\u8fed\u4ee3\u8fd9\u5957\u80fd\u529b\u7275\u5f15\u7684\u6570\u636e\u5f15\u64ce\u3002<\/p>\n<p class=\"tdx-end\">\u00b7 END \u00b7<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u673a\u5668\u4eba\u5206\u4e0d\u6e05\u753b\u9762\u91cc&#8221;\u66f4\u9760\u4e0a&#8221;\u548c\u771f\u5b9e\u4e09\u7ef4\u7a7a\u95f4&#8221;\u66f4\u9ad8&#8221;\u7684\u5dee\u522b\uff0c\u89c6\u89d2\u4e00\u6362\u5c31\u4f1a\u8bef\u5224\u7269\u4f53\u8eab\u4efd\u3001\u76f8\u673a\u8fd0\u52a8\u6216\u4efb\u52a1\u8fdb\u5ea6\u3002\u62db\u5546\u5c40\u5148\u8fdb\u6280\u672f\u7814\u7a76\u9662\u4e0b\u5c5e\u72ee\u5b50\u5c71\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u8fd1\u65e5\u53d1\u5e03\u5177\u8eab\u63a8\u7406\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\u300c\u9a6d\u7269\u2122 Lion-ER-35B-A3B\u300d\uff0c\u603b\u53c2\u6570\u7ea635B\uff0c\u6bcf\u4e2aToken\u4ec5\u6fc0\u6d3b\u7ea63B\u53c2\u6570\uff0c\u5728\u5341\u9879\u4e3b\u6d41\u5177\u8eab\u76f8\u5173\u57fa\u51c6\u4e0a\u5168\u90e8\u6307\u6807\u4e0a\u6da8\uff0c\u7b49\u6743\u7efc\u5408\u5747\u5206\u8fbe\u2026<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[],"class_list":["post-1104","post","type-post","status-publish","format-standard","hentry","category-4"],"_links":{"self":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/posts\/1104","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/comments?post=1104"}],"version-history":[{"count":0,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/posts\/1104\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/media?parent=1104"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/categories?post=1104"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/tags?post=1104"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}