{"id":12562,"date":"2026-09-23T18:20:56","date_gmt":"2026-09-23T10:20:56","guid":{"rendered":"https:\/\/infernews.com\/blog\/taste-bench-measuring-the-long-horizon-judgment-of-llm-agents-at-real-decision\/"},"modified":"2026-09-23T18:20:56","modified_gmt":"2026-09-23T10:20:56","slug":"taste-bench-measuring-the-long-horizon-judgment-of-llm-agents-at-real-decision","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/taste-bench-measuring-the-long-horizon-judgment-of-llm-agents-at-real-decision\/","title":{"rendered":"Taste-Bench \uff1a\u907f\u514d\u628a\u6642\u9593\u548c\u8a08\u7b97\u8cc7\u6e90\u82b1\u5728\u932f\u8def\u4e0a"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/logo-2.jpg\" alt=\"Taste-Bench\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">AI Agent \u505a\u9577\u4efb\u52d9\u6642\uff0c\u9078\u932f\u65b9\u5411\u53ef\u80fd\u8981\u5230\u591a\u500b\u6b65\u9a5f\u4e4b\u5f8c\u624d\u770b\u5f97\u51fa\u4ee3\u50f9\u3002Taste-Bench \u662f\u4e00\u5957\u8a55\u6e2c\u8cc7\u6599\u96c6\uff0c\u8b93\u6a21\u578b\u95b1\u8b80\u4efb\u52d9\u80cc\u666f\u3001\u5206\u5c94\u524d\u7684\u64cd\u4f5c\u7d00\u9304\uff0c\u4ee5\u53ca\u5169\u500b\u5019\u9078\u4e0b\u4e00\u6b65\uff0c\u518d\u5224\u65b7\u54ea\u500b\u9078\u64c7\u8f03\u597d\uff1b\u5b83\u6e2c\u8a66\u7684\u4e0d\u662f\u55ae\u6b65\u7b54\u984c\uff0c\u800c\u662f\u80fd\u5426\u907f\u514d\u628a\u6642\u9593\u548c\u8a08\u7b97\u8cc7\u6e90\u82b1\u5728\u932f\u8def\u4e0a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8cc7\u6599\u96c6\u6709 502 \u984c\uff0c\u5f9e\u8edf\u4ef6\u5de5\u7a0b\u8207\u6a5f\u5668\u5b78\u7fd2\u7814\u7a76\u4efb\u52d9\u7684\u64cd\u4f5c\u8ecc\u8de1\u4e2d\u6574\u7406\u800c\u6210\uff0c\u6c92\u6709\u9010\u984c\u7531\u5c08\u5bb6\u6a19\u8a3b\u3002\u984c\u76ee\u5206\u70ba\u5169\u985e\uff1a\u4e0d\u540c\u5617\u8a66\u5728\u540c\u4e00\u4f4d\u7f6e\u5206\u5c94\uff0c\u6700\u5f8c\u7d50\u679c\u63d0\u4f9b\u5224\u65b7\u4f9d\u64da\uff1b\u6216\u540c\u4e00\u6b21\u57f7\u884c\u4e2d\uff0cAgent \u9047\u5230\u5931\u6557\u5f8c\u653e\u68c4\u539f\u8def\u4e26\u6062\u5fa9\u3002\u7814\u7a76\u5718\u968a\u5f9e 4,657 \u500b\u5019\u9078\u5206\u5c94\u4e2d\u5254\u9664\u904e\u65bc\u7c21\u55ae\u6216\u7121\u6cd5\u5224\u5b9a\u7684\u984c\u76ee\uff0c\u4e26\u4ee5\u5f8c\u7e8c\u7d50\u679c\u6a19\u793a\u8f03\u4f73\u65b9\u5411\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6e2c\u8a66\u6642\u53ef\u5f9e Hugging Face \u53d6\u5f97\u8cc7\u6599\u96c6\uff0c\u6309\u984c\u76ee\u63d0\u4f9b\u7684\u7d00\u9304\u6bd4\u8f03\u6a21\u578b\u9078\u64c7\uff1b\u7a0b\u5f0f\u78bc\u5eab\u4ea6\u63d0\u4f9b Python 3.11 \u4ee5\u4e0a\u7684\u5feb\u901f\u958b\u59cb\u65b9\u5f0f\u3002\u8a55\u5206\u8981\u6c42\u6a21\u578b\u5728\u5169\u7a2e\u9078\u9805\u6392\u5217\u9806\u5e8f\u4e0b\u90fd\u7b54\u5c0d\uff0c\u6e1b\u4f4e\u55ae\u9760\u56fa\u5b9a\u9078\u4f4d\u7f6e\u53d6\u5f97\u9ad8\u5206\u7684\u53ef\u80fd\u3002\u8f03\u8cbc\u8fd1\u65b9\u5411\u5224\u65b7\u80fd\u529b\uff0c\u4f46\u984c\u76ee\u4ecd\u96c6\u4e2d\u65bc SWE-bench\u3001SWE-bench Pro\uff0c\u4ee5\u53ca METR \u7684 MALT \u7814\u7a76\u8ecc\u8de1\uff0c\u672a\u5fc5\u6db5\u84cb\u6240\u6709\u9577\u4efb\u52d9\u5834\u666f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">14 \u500b\u524d\u6cbf\u6a21\u578b\u4e2d\uff0c\u6700\u9ad8\u5e73\u5747\u6b63\u78ba\u7387\u70ba 59.7%\uff0c\u4ecd\u6709\u7d04\u56db\u6210\u984c\u76ee\u5224\u65b7\u932f\u8aa4\uff1b\u7576\u95dc\u9375\u8b49\u64da\u8981\u5230\u66f4\u5f8c\u6bb5\u624d\u51fa\u73fe\uff0c\u6e96\u78ba\u7387\u6703\u7531 62.3% \u964d\u81f3 21.0%\uff0c\u589e\u52a0\u63a8\u7406\u9810\u7b97\u4ea6\u672a\u80fd\u6539\u5584\u7d50\u679c\u3002\u7814\u7a76\u4ea6\u5831\u544a\uff0c\u900f\u904e\u4e8b\u5f8c\u7b54\u6848\u84b8\u993e\u8a13\u7df4\u7684 Qwen3.6-27B\uff0c\u5728\u672a\u898b\u4efb\u52d9\u4e0a\u7684\u6b63\u78ba\u7387\u7531 30.0% \u5347\u81f3 47.9%\uff1b\u7528\u5b83\u63d0\u4f9b\u5efa\u8b70\u5f8c\uff0cSWE-bench Pro \u6210\u529f\u7387\u7531 14.6% \u5347\u81f3 33.7%\u3002\u9019\u4e9b\u7d50\u679c\u986f\u793a\u8a55\u6e2c\u4e0d\u53ea\u7528\u4f86\u6bd4\u8f03\u6a21\u578b\uff0c\u4e5f\u63d0\u4f9b\u4e86\u6539\u5584 Agent \u6c7a\u7b56\u7684\u8a13\u7df4\u65b9\u5411\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u8a55\u6e2c\u6838\u5fc3\uff1a\u5728\u7d50\u679c\u5c1a\u672a\u63ed\u66c9\u6642\uff0c\u5f9e\u5169\u500b\u4e0b\u4e00\u6b65\u4e2d\u9078\u51fa\u8f03\u4f73\u65b9\u5411\u3002<\/li>\n<li>\u984c\u76ee\u4f86\u6e90\uff1a\u8edf\u4ef6\u5de5\u7a0b\u8207\u6a5f\u5668\u5b78\u7fd2\u7814\u7a76\u4efb\u52d9\u7684\u771f\u5be6\u64cd\u4f5c\u8ecc\u8de1\u3002<\/li>\n<li>\u4e3b\u8981\u9650\u5236\uff1a\u76ee\u524d 502 \u984c\u7684\u9818\u57df\u8207\u4efb\u52d9\u4f86\u6e90\u6709\u9650\uff0c\u4e0d\u80fd\u4ee3\u8868\u6240\u6709 Agent \u5de5\u4f5c\u3002<\/li>\n<li>\u9069\u5408\u5c0d\u8c61\uff1a\u7814\u7a76\u9577\u4efb\u52d9 Agent\u3001\u6bd4\u8f03\u6a21\u578b\u6c7a\u7b56\u80fd\u529b\uff0c\u6216\u8a13\u7df4\u5de5\u5177\u4f7f\u7528\u7b56\u7565\u7684\u5718\u968a\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/huggingface.co\/papers\/2609.25804\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/wbopan\/tastebench\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>AI Agent \u8981\u505a\u7684\u9078\u64c7\uff0c\u7576\u4e0b\u5f80\u5f80\u770b\u4f3c\u5408\u7406\uff0c\u4ee3\u50f9\u537b\u53ef\u80fd\u5728\u5f88\u4e45\u4e4b\u5f8c\u624d\u6d6e\u73fe\u3002Taste-Bench \u628a\u9019\u985e\u6c7a\u7b56\u5206\u5c94\u6574\u7406\u6210\u6e2c\u8a66\uff0c\u6aa2\u67e5\u6a21\u578b\u80fd\u5426\u53ca\u65e9\u9078\u5c0d\u65b9\u5411\u3002<\/p>\n","protected":false},"author":8,"featured_media":12561,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,116,127,185,38,152,199],"tags":[],"class_list":["post-12562","post","type-post","status-publish","format-standard","hentry","category-133","category-agentic","category-127","category-qwen","category-38","category-python","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12562","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=12562"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12562\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/12561"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=12562"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=12562"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=12562"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}