
{"id":8709,"date":"2026-06-02T04:13:31","date_gmt":"2026-06-01T20:13:31","guid":{"rendered":"https:\/\/infernews.com\/blog\/rots\/"},"modified":"2026-06-02T04:13:31","modified_gmt":"2026-06-01T20:13:31","slug":"rots","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/rots\/","title":{"rendered":"GUI \u667a\u80fd\u9ad4\u7e3d\u662f\u81ea\u5df1\u72af\u932f\uff1fRoTS \u7528\u6a39\u72c0\u8ecc\u8de1\u5408\u6210\u6559\u5b83\u5982\u4f55\u81ea\u6551"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pasted-81626d0995b9.jpg\" alt=\"Repository image for AlibabaResearch\/RoTS\"><\/figure>\n<p>\u8fd1\u5e74 GUI agents\uff08\u5716\u5f62\u4ecb\u9762\u667a\u80fd\u9ad4\uff09\u96d6\u7136\u9032\u6b65\u795e\u901f\uff0c\u4f46\u53ea\u8981\u81ea\u5df1\u9ede\u932f\u4e00\u500b\u6309\u9215\u3001\u8aa4\u5224\u756b\u9762\u72c0\u614b\uff0c\u5f80\u5f80\u5c31\u6703\u5361\u4f4f\u7121\u6cd5\u5b8c\u6210\u4efb\u52d9\u3002\u9019\u9805\u7531\u963f\u91cc\u96f2\u5718\u968a\u64b0\u5beb\u3001\u5165\u9078 ICML 2026 Spotlight \u7684\u5de5\u4f5c\uff0c\u6b63\u662f\u91dd\u5c0d\u9019\u500b\u300c\u81ea\u5df1\u9020\u6210\u7684\u932f\u8aa4\u300d\u75db\u9ede\uff0c\u5f9e\u8a55\u4f30\u8207\u8a13\u7df4\u6578\u64da\u5169\u7aef\u540c\u6642\u5165\u624b\u3002<\/p>\n<p>\u4ed6\u5011\u5148\u63a8\u51fa\u4e86 GUI-RobustEval \u57fa\u6e96\uff0c\u6536\u9304 1,216 \u500b\u53ef\u57f7\u884c\u6e2c\u8a66\u6848\u4f8b\uff0c\u6db5\u84cb 11 \u7a2e\u932f\u8aa4\u985e\u578b\u8207 4 \u7a2e\u932f\u8aa4\u6df1\u5ea6\uff0c\u8b93\u958b\u767c\u8005\u80fd\u7cfb\u7d71\u5316\u91cf\u6e2c\u667a\u80fd\u9ad4\u7684\u300c\u81ea\u6551\u300d\u80fd\u529b\u3002\u63a5\u8457\u63d0\u51fa RoTS \u6846\u67b6\uff0c\u4ee5\u6a39\u72c0\u7d50\u69cb\uff08tree-based\uff09\u5728\u7dda\u751f\u6210 80 \u842c\u689d\u8a13\u7df4\u8ecc\u8de1\uff0c\u4e3b\u52d5\u63a2\u7d22\u4e0d\u540c\u932f\u8aa4\u6a21\u5f0f\u4e26\u5408\u6210\u5c0d\u61c9\u7684\u6062\u5fa9\u6b65\u9a5f\u3002<\/p>\n<p>\u57fa\u65bc\u9019\u6279\u6578\u64da\u5fae\u8abf\u800c\u6210\u7684 RoTS-7B \u8207 RoTS-32B\uff0c\u5728 OSWorld \u4e0a\u65bc\u958b\u6e90\u6b0a\u91cd\u6a21\u578b\u4e2d\u53d6\u5f97\u9818\u5148\u8868\u73fe\uff0c\u5176\u4e2d RoTS-32B \u9054\u5230 47.4% \u6210\u529f\u7387\u8207 33.8% All-Pass@4 \u5206\u6578\u3002\u7814\u7a76\u5718\u968a\u6307\u51fa\uff0c\u9019\u4e9b\u5206\u6578\u986f\u793a\u9577\u6642\u7a0b\u932f\u8aa4\u6062\u5fa9\u80fd\u529b\uff0c\u540c\u6642\u63d0\u5347\u4e86\u6574\u9ad4\u4efb\u52d9\u8868\u73fe\u3002<\/p>\n<p>\u5c0d\u5f9e\u4e8b\u96fb\u8166\u4f7f\u7528\u4ee3\u7406\uff08Computer-use agents, CUAs\uff09\u7814\u7a76\u3001\u958b\u6e90 VLM \u5fae\u8abf\uff0c\u6216\u95dc\u6ce8 GUI \u81ea\u52d5\u5316\u5728\u771f\u5be6\u74b0\u5883\u7a69\u5b9a\u6027\u7684\u5718\u968a\u800c\u8a00\uff0c\u9019\u662f\u4e00\u4efd\u503c\u5f97\u8ffd\u8e64\u7684\u6210\u679c\u3002\u8ad6\u6587\u8207\u6578\u64da\u96c6\u5df2\u516c\u958b\uff0c\u4f46\u7a0b\u5f0f\u78bc\u8207\u8a55\u4f30\u5de5\u5177\u4ecd\u5728\u6574\u7406\u4e2d\uff0c\u6709\u8208\u8da3\u7684\u4eba\u53ef\u5148\u95b1\u8b80\u8ad6\u6587\u4e26\u95dc\u6ce8\u5f8c\u7e8c\u91cb\u51fa\u3002<\/p>\n<p>\u91cd\u9ede\u6458\u8981\uff1a<\/p>\n<ul>\n<li>\u89e3\u6c7a GUI agents \u56e0\u81ea\u8eab\u7b56\u7565\u932f\u8aa4\u800c\u7121\u6cd5\u6062\u5fa9\u7684\u90e8\u7f72\u74f6\u9838<\/li>\n<li>GUI-RobustEval \u63d0\u4f9b 1,216 \u500b\u6e2c\u8a66\u6848\u4f8b\uff0c\u8986\u84cb 11 \u7a2e\u932f\u8aa4\u985e\u578b<\/li>\n<li>RoTS \u4ee5\u6a39\u72c0\u5728\u7dda\u5408\u6210\u6846\u67b6\u7522\u51fa 80 \u842c\u689d\u9ad8\u54c1\u8cea\u8a13\u7df4\u8ecc\u8de1<\/li>\n<li>RoTS-7B \u8207 RoTS-32B \u65bc OSWorld \u958b\u6e90\u6a21\u578b\u4e2d\u8868\u73fe\u9818\u5148<\/li>\n<li>\u7a0b\u5f0f\u78bc\u8207\u6578\u64da\u96c6\u4ecd\u5728\u6574\u7406\u968e\u6bb5\uff0c\u8ad6\u6587\u5df2\u65bc arXiv \u516c\u958b<\/li>\n<\/ul>\n<p><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/AlibabaResearch\/RoTS\" rel=\"noopener noreferrer\">https:\/\/github.com\/AlibabaResearch\/RoTS<\/a><\/p>\n<p><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2605.29447\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2605.29447<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u963f\u91cc\u96f2\u767c\u5e03\u7684 RoTS \u9805\u76ee\uff0c\u70ba GUI \u667a\u80fd\u9ad4\u88dc\u4e0a\u300c\u72af\u932f\u5f8c\u6062\u5fa9\u300d\u7684\u80fd\u529b\u77ed\u677f\uff0c\u4e26\u5728 OSWorld \u57fa\u6e96\u4e0a\u5237\u65b0\u958b\u6e90\u6a21\u578b\u8868\u73fe\u3002<\/p>\n","protected":false},"author":8,"featured_media":8708,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[133,116,190,197],"tags":[],"class_list":["post-8709","post","type-post","status-publish","format-standard","hentry","category-133","category-agentic","category-openclaw","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8709","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=8709"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8709\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8708"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8709"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8709"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8709"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}