{"id":1036,"date":"2026-10-30T21:30:21","date_gmt":"2026-10-30T13:30:21","guid":{"rendered":"https:\/\/www.timedigital.cn\/new\/1036\/"},"modified":"2026-10-30T21:30:21","modified_gmt":"2026-10-30T13:30:21","slug":"navital-gym%ef%bc%9a%e5%8d%b0%e5%ba%a6%e5%9c%ba%e6%99%af-ai-agent-%e6%b5%8b%e8%af%95%e9%9b%86","status":"publish","type":"post","link":"https:\/\/www.timedigital.cn\/new\/1036\/","title":{"rendered":"Navital Gym\uff1a\u5370\u5ea6\u573a\u666f AI Agent \u6d4b\u8bd5\u96c6"},"content":{"rendered":"<div class=\"tdx-art\">\n<p class=\"tdx-lead\">Navital Labs \u53d1\u5e03 Navital Gym\uff0c\u4e00\u5957\u9762\u5411\u5370\u5ea6\u672c\u571f\u4efb\u52a1\u7684 AI agent \u5f00\u6e90\u6d4b\u8bd5\u73af\u5883\uff0c\u4efb\u52a1\u7531\u786e\u5b9a\u6027\u89c4\u5219\u81ea\u52a8\u8bc4\u5206\uff0c\u4e0d\u5f15\u5165 AI \u88c1\u5224\uff0c\u9996\u8f6e\u57fa\u7ebf\u8dd1\u5206\u5df2\u516c\u5e03\u5728\u9879\u76ee\u7684 GitHub \u4ed3\u5e93\u3002\u6d4b\u8bd5\u96c6\u8986\u76d6 6 \u7c7b\u4efb\u52a1\uff0c\u4ee3\u7801\u5904\u4e8e\u6d3b\u8dc3\u5f00\u53d1\u671f\uff0c\u8dd1\u5206\u7ed3\u679c\u968f\u5b8c\u6210\u7684\u6279\u6b21\u8865\u8fdb\u6587\u6863\u3002<\/p>\n<p>\u516d\u7c7b\u4efb\u52a1\u5206\u522b\u662f\uff1a<\/p>\n<ul>\n<li>trip\uff1a\u5370\u5ea6\u706b\u8f66\u914d\u989d\u3001\u5019\u8865\u540d\u5355\u3001\u822a\u73ed\u3001\u516c\u8def\u3001\u9152\u5e97\u3001\u5b63\u8282\u548c\u8bb8\u53ef\u8bc1\u7ea6\u675f\u4e0b\u7684\u884c\u7a0b\u89c4\u5212\uff0c55 \u4e2a\u4efb\u52a1\uff0c\u5176\u4e2d 17 \u4e2a\u65e0\u53ef\u884c\u89e3<\/li>\n<li>gst\uff1aGST \u53d1\u7968\u5f00\u5177\u4e0e\u7533\u62a5\u6570\u989d\u8ba1\u7b97\uff0c50 \u4e2a\u4efb\u52a1<\/li>\n<li>form\uff1a\u8868\u5355\u586b\u5199\u4e0e\u6821\u9a8c\uff0c50 \u4e2a\u4efb\u52a1<\/li>\n<li>support\uff1a\u5370\u5730\u8bed\u4e0e\u5370\u82f1\u6df7\u7528\u573a\u666f\u7684\u9000\u8d27\u3001\u9000\u6b3e\u3001\u9690\u79c1\u8bf7\u6c42\u5ba2\u670d\uff0c50 \u4e2a\u6838\u5fc3\u4efb\u52a1\u548c 26 \u4e2a\u5b9e\u9a8c\u4efb\u52a1<\/li>\n<li>form_multidoc\uff1a\u8de8\u8eab\u4efd\u8bc1\u4ef6\u548c\u8bb0\u5f55\u6bd4\u5bf9\u3001\u767b\u8bb0\u51b2\u7a81\u4e0e OCR \u4fee\u590d\u7b49\uff0c40 \u4e2a\u4efb\u52a1<\/li>\n<li>upi\uff1aUPI \u652f\u4ed8\u5c1d\u8bd5\u4e0e\u7ea0\u7eb7\u5904\u7406\uff0c50 \u4e2a\u4efb\u52a1<\/li>\n<\/ul>\n<p>\u6bcf\u5957\u73af\u5883\u4f7f\u7528\u7b80\u5316\u7248\u7684\u5370\u5ea6\u5b9e\u52a1\u89c4\u5219\u548c\u5408\u6210\u6570\u636e\uff0c\u73af\u5883 README \u5199\u660e\u89c4\u5219\u51fa\u5904\u548c\u7248\u672c\u65e5\u671f\u3002\u8bc4\u5206\u662f\u786e\u5b9a\u6027\u6267\u884c\u7684\uff1aagent \u58f0\u79f0\u7684\u6210\u672c\u3001\u5f97\u5206\u548c\u201c\u5df2\u5b8c\u6210\u201d\u8868\u8ff0\u4e00\u5f8b\u4e0d\u91c7\u4fe1\uff0c\u5783\u573e\u8f93\u5165\u5f97 0 \u5206\uff0c\u8bc4\u5206\u5668\u4e0d\u4f1a\u5d29\u6e83\u3002\u4e0d\u53ef\u884c\u4efb\u52a1\u662f\u6d4b\u8bd5\u7684\u4e00\u90e8\u5206\uff0cagent \u9700\u8981\u660e\u786e\u8bf4\u660e\u4e0d\u53ef\u884c\u5e76\u7ed9\u51fa\u7406\u7531\uff0c\u7406\u7531\u9519\u8bef\u5373\u6263\u5206\u3002\u9884\u671f\u7b54\u6848\u3001\u53ef\u884c\u6027\u5224\u5b9a\u548c\u8bbe\u8ba1\u5907\u6ce8\u7559\u5728\u73af\u5883\u5185\u90e8\uff0cagent \u53ea\u80fd\u62ff\u5230\u63d0\u793a\u3001\u5de5\u5177\u548c\u89c4\u5219\u6587\u672c\uff0c\u5931\u8d25\u6d4b\u8bd5\u5f3a\u5236\u8fd9\u4e00\u9694\u79bb\u3002\u6838\u5fc3\u548c\u5404\u73af\u5883\u53ea\u5728 Python 3.10+ \u6807\u51c6\u5e93\u4e0a\u8fd0\u884c\uff0cInspect AI\u3001verifiers \u548c OpenEnv \u9002\u914d\u5668\u4e3a\u53ef\u9009\u7ec4\u4ef6\u3002<\/p>\n<p><strong>claude-haiku-4-5 \u5728 40 \u4e2a multi-doc \u4efb\u52a1\u4e0a pass@1 \u4e3a 0.98<\/strong>\uff0cform \u4e0a\u4e3a 0.82\uff0c55 \u4e2a trip \u4efb\u52a1\u4e0a\u4e3a 0.31\uff0c76 \u4e2a support \u4efb\u52a1\u4e0a\u4e3a 0.24\u3002gpt-4o-mini \u5728 form \u4e0a 0.54\uff0ctrip \u4e0a <strong>\u53ea\u6709 0.05<\/strong>\uff0csupport \u4e0a 0.09\u3002\u8868\u5185\u5355\u72ec\u5217\u51fa easy \u96be\u5ea6\u6863\uff0cclaude-haiku-4-5 \u7684 gst easy \u62ff\u5230 1.00\uff0ctrip easy \u4e3a 0.67\uff1b\u8fd9\u4e00\u6863\u7684\u8bbe\u8ba1\u76ee\u6807\u662f\u8ba9 7B \u5230 30B \u91cf\u7ea7\u7684\u6a21\u578b\u5076\u5c14\u89e3\u51fa\u90e8\u5206\u4efb\u52a1\u3002<\/p>\n<p>\u4e09\u5957\u9002\u914d\u5668\u90fd\u5e26\u79bb\u7ebf\u7aef\u5230\u7aef\u6d4b\u8bd5\u3002verifiers 0.4.0 v1 \u4e0e openenv-core 0.3.0 \u63a5\u53e3\u8f83\u65b0\uff0c\u9879\u76ee\u65b9\u5199\u660e\u8fd9\u4e9b\u9002\u914d\u5668\u5c1a\u672a\u5728\u5bb9\u5668\u6216\u6258\u7ba1\u7a7a\u95f4\u5185\u5b9e\u9645\u8fd0\u884c\u3002\u9879\u76ee\u4ee5 v1.0.1 \u7248\u672c\u53f7\u56fa\u5b9a\u8bc4\u5206\u4ee3\u7801\uff0cdocs\/RESULTS.md \u7684\u7ed3\u679c\u8868\u7531 scripts\/regrade.py \u91cd\u65b0\u8bc4\u5206\u540e\u968f\u65b0\u5b8c\u6210\u6279\u6b21\u66f4\u65b0\u3002<\/p>\n<p class=\"tdx-end\">\u00b7 END \u00b7<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>Navital Labs \u53d1\u5e03 Navital Gym\uff0c\u4e00\u5957\u9762\u5411\u5370\u5ea6\u672c\u571f\u4efb\u52a1\u7684 AI agent \u5f00\u6e90\u6d4b\u8bd5\u73af\u5883\uff0c\u4efb\u52a1\u7531\u786e\u5b9a\u6027\u89c4\u5219\u81ea\u52a8\u8bc4\u5206\uff0c\u4e0d\u5f15\u5165 AI \u88c1\u5224\uff0c\u9996\u8f6e\u57fa\u7ebf\u8dd1\u5206\u5df2\u516c\u5e03\u5728\u9879\u76ee\u7684 GitHub \u4ed3\u5e93\u3002\u6d4b\u8bd5\u96c6\u8986\u76d6 6 \u7c7b\u4efb\u52a1\uff0c\u4ee3\u7801\u5904\u4e8e\u6d3b\u8dc3\u5f00\u53d1\u671f\uff0c\u8dd1\u5206\u7ed3\u679c\u968f\u5b8c\u6210\u7684\u6279\u6b21\u8865\u8fdb\u6587\u6863\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[],"class_list":["post-1036","post","type-post","status-publish","format-standard","hentry","category-3"],"_links":{"self":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/posts\/1036","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=1036"}],"version-history":[{"count":0,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/posts\/1036\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/media?parent=1036"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/categories?post=1036"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.timedigital.cn\/new\/wp-json\/wp\/v2\/tags?post=1036"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}