{"id":11289,"date":"2026-08-21T17:02:31","date_gmt":"2026-08-21T09:02:31","guid":{"rendered":"https:\/\/infernews.com\/blog\/envharness\/"},"modified":"2026-08-21T17:13:04","modified_gmt":"2026-08-21T09:13:04","slug":"envharness","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/envharness\/","title":{"rendered":"EnvHarness \u8b93\u975c\u614b\u74b0\u5883\u6309\u4ee3\u7406\u5f31\u9ede\u6301\u7e8c\u9032\u5316"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/example.jpg\" alt=\"example\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Google Cloud AI Research \u806f\u540c Washington University in St. Louis\u3001Google Cloud\u3001University of North Carolina at Chapel Hill \u7684\u7814\u7a76\u4eba\u54e1\uff0c\u958b\u767c\u4e86 EnvHarness\u3002\u9019\u500b\u958b\u6e90\u6846\u67b6\u8655\u7406\u7684\u662f\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08Large Language Model\uff0cLLM\uff09\u4ee3\u7406\u9762\u5c0d\u975c\u614b\u4e92\u52d5\u74b0\u5883\u6642\uff0c\u4efb\u52d9\u548c\u56de\u994b\u7121\u6cd5\u96a8\u80fd\u529b\u8b8a\u5316\u7684\u554f\u984c\uff1b\u5b83\u4fdd\u7559\u539f\u6709\u74b0\u5883\u53ca\u9a57\u8b49\u5668\uff0c\u6539\u7531\u5916\u52a0\u5c64\u63a7\u5236\u4ee3\u7406\u6240\u898b\u3001\u53ef\u505a\u7684\u884c\u52d5\u548c\u8d77\u59cb\u72c0\u614b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4ee3\u7406\u5728\u540c\u4e00\u6279\u4efb\u52d9\u4e0a\u53cd\u8986\u7df4\u7fd2\uff0c\u5f80\u5f80\u6703\u5f88\u5feb\u89e3\u5b8c\uff0c\u74b0\u5883\u537b\u4e0d\u80fd\u56e0\u61c9\u5f31\u9ede\u63d0\u4f9b\u65b0\u6311\u6230\u3002EnvHarness \u662f\u4e00\u500b\u9762\u5411\u4ee3\u7406\u5b78\u7fd2\u7684\u958b\u6e90\u6846\u67b6\uff0c\u900f\u904e\u6a19\u6e96 <code>reset<\/code>\uff0f<code>step<\/code> \u4ecb\u9762\u5305\u88f9\u56fa\u5b9a\u74b0\u5883\uff0c\u8abf\u6574\u4ee3\u7406\u7684\u8d77\u59cb\u72c0\u614b\u3001\u53ef\u7528\u884c\u52d5\u8207\u89c0\u5bdf\u5167\u5bb9\uff0c\u540c\u6642\u4fdd\u7559\u539f\u6709\u7684\u6210\u529f\u5224\u5b9a\u5668\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u628a\u63a7\u5236\u62c6\u6210\u4e09\u985e\u53ef\u7d44\u5408\u5143\u4ef6\uff1aSetup \u8ca0\u8cac\u91cd\u5851\u521d\u59cb\u72c0\u614b\uff0cRule \u6539\u8b8a\u4e92\u52d5\u898f\u5247\u3001\u884c\u52d5\u9650\u5236\u53ca\u56de\u994b\uff0cLink \u5247\u628a\u53e6\u4e00\u500b\u74b0\u5883\u7684\u4efb\u52d9\u63a5\u5165\u76ee\u524d\u6d41\u7a0b\u3002\u8a2d\u8a08\u8005\u4ee3\u7406\u6703\u8b80\u53d6\u76ee\u6a19\u4ee3\u7406\u7684\u64cd\u4f5c\u8ecc\u8de1\uff0c\u8a3a\u65b7\u5931\u8aa4\uff0c\u518d\u4ee5 Python \u5beb\u51fa\u5143\u4ef6\u3001\u6e2c\u8a66\u53ca\u4fee\u8a02\uff0c\u4ee4\u8a13\u7df4\u74b0\u5883\u96a8\u4ee3\u7406\u80fd\u529b\u4e00\u540c\u8abf\u6574\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u2022 \u4e0d\u9700\u4fee\u6539 ALFWorld\u3001WebArena\u3001SWE-bench Verified \u7b49\u74b0\u5883\u7684\u5167\u90e8\u7a0b\u5f0f\u78bc<br>\n\u2022 \u4efb\u52d9\u8207 verifier \u4fdd\u6301\u4f86\u6e90 benchmark \u7684\u53ef\u4fe1\u5224\u5b9a<br>\n\u2022 \u5143\u4ef6\u53ef\u81ea\u7531\u5806\u758a\uff0c\u9069\u5408\u91dd\u5c0d\u7279\u5b9a\u5f31\u9ede\u6539\u9020\u4e92\u52d5\u6d41\u7a0b<br>\n\u2022 \u53ef\u5728\u9805\u76ee\u63d0\u4f9b\u7684\u700f\u89bd\u5668 Playground \u4ee5 Toy24Env \u89c0\u5bdf\u74b0\u5883\u72c0\u614b\u8207\u4ee3\u7406\u8996\u89d2<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5728 ALFWorld\u3001WebArena \u53ca SWE-bench Verified \u7684 EnvHarness \u8a13\u7df4\u7d50\u679c\u5206\u5225\u9054\u5230 68.3\u300141.6 \u53ca 52.6\uff0c\u5c55\u793a\u76f8\u5c0d\u57fa\u6e96\u6700\u9ad8 5.9 \u500b\u767e\u5206\u9ede\u7684\u63d0\u5347\u3002\u4e0d\u904e\uff0c\u6548\u679c\u53d6\u6c7a\u65bc\u8a2d\u8a08\u8005\u4ee3\u7406\u80fd\u5426\u6e96\u78ba\u8a3a\u65b7\u8ecc\u8de1\uff0c\u4ee5\u53ca\u6539\u5beb\u5f8c\u7684\u74b0\u5883\u662f\u5426\u771f\u7684\u5c0d\u61c9\u76ee\u6a19\u5f31\u9ede\uff0c\u4e26\u975e\u55ae\u9760\u5957\u4e0a\u5143\u4ef6\u5c31\u80fd\u4fdd\u8b49\u6539\u5584\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u53d6\u6368\u662f\u628a\u74b0\u5883\u8a2d\u8a08\u5de5\u4f5c\u8f49\u79fb\u5230 Setup\u3001Rule\u3001Link \u7684\u7a0b\u5f0f\u7de8\u5beb\u8207\u53cd\u8986\u6e2c\u8a66\uff0c\u63db\u4f86\u8de8 benchmark \u91cd\u7528\u548c\u8f03\u4f4e\u7684\u74b0\u5883\u6539\u9020\u6210\u672c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/envharness.com\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/google-research\/envharness\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>EnvHarness \u4e0d\u6539\u52d5\u74b0\u5883\u5167\u90e8\u7a0b\u5f0f\u78bc\uff0c\u900f\u904e\u53ef\u5806\u758a\u5143\u4ef6\u8abf\u6574\u4efb\u52d9\u3001\u898f\u5247\u8207\u89c0\u5bdf\uff0c\u8b93\u4ee3\u7406\u7372\u5f97\u66f4\u6709\u91dd\u5c0d\u6027\u7684\u8a13\u7df4\u3002<\/p>\n","protected":false},"author":8,"featured_media":11288,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,178,116,152,199],"tags":[],"class_list":["post-11289","post","type-post","status-publish","format-standard","hentry","category-133","category-google","category-agentic","category-python","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11289","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=11289"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11289\/revisions"}],"predecessor-version":[{"id":11292,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11289\/revisions\/11292"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/11288"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=11289"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=11289"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=11289"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}