{"id":6530,"date":"2025-04-07T06:46:00","date_gmt":"2025-04-06T21:46:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6530"},"modified":"2025-04-07T06:46:00","modified_gmt":"2025-04-06T21:46:00","slug":"agent-s2-devin-2-amazon-nova-act","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6530","title":{"rendered":"Agent S2, Devin 2,\u00a0Amazon Nova Act, An Illusion of Progress? Assessing the Current State of Web Agents\u00a0"},"content":{"rendered":"\n<p>\u4ee5\u524d\u53d6\u308a\u4e0a\u3052\u305fAgent S\u306e\u30d0\u30fc\u30b8\u30e7\u30f32\u304c\u51fa\u3066\u3044\u305f\u3002\u534a\u5e74\u3067OS World\u306e\u30b9\u30b3\u30a2\u304c20.5\u304b\u308927.0\uff0815Step\uff09\u306b\u4e0a\u304c\u3063\u3066\u304a\u308a\u3001\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\uff08LLM\uff09\u306e\u6027\u80fd\u5411\u4e0a\u3082\u3042\u308b\u3060\u308d\u3046\u304c\u7740\u5b9f\u306a\u9032\u5316\u3092\u611f\u3058\u308b\u3002<a href=\"https:\/\/labs.amazon.science\/blog\/nova-act\">Introducing Amazon Nova Act | Amazon AGI Labs<\/a>\u3001<a href=\"https:\/\/cognition.ai\/blog\/devin-2\">Cognition | Devin 2.0<\/a>\u306a\u3069\u767a\u8868\u304c\u76f8\u6b21\u304eGUI Agent\u7684\u306aLLM based Agent\u306f\u6d41\u884c\u3057\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<p>\u500b\u4eba\u306e\u30b5\u30a4\u30c8\u3067\u3082fugumt.com\u306f<a href=\"https:\/\/fugumt.com\/fugu_agent\/top\">Fugu-MT:Agent<\/a>\u3067Agent\u5316\u3092\u884c\u3063\u3066\u3044\u308b\uff08<a href=\"https:\/\/staka.jp\/wordpress\/?p=1110\">OpenManus\u3092\u4f7f\u3063\u305f\u30b5\u30a4\u30c8\u3078\u306e\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u7d44\u307f\u8fbc\u307f | \u3077\u308b\u30fc\u3075\u304a\u3076\u3053\u3093\u305b\u3077\u3068<\/a>\uff09\u3002\u5bb9\u6613\u306b\u30b5\u30a4\u30c8\u306e\u6a5f\u80fd\u3092\u62e1\u5f35\u3067\u304d\u308b\u3053\u3068\u304b\u3089\u3001\u4eca\u5f8c\u3053\u306e\u3088\u3046\u306a\u30b5\u30a4\u30c8\u304c\u5897\u3048\u3066\u304f\u308b\u306e\u3067\u306f\u306a\u3044\u304b\u3068\u601d\u3046(*1)\u3002<\/p>\n\n\n\n<p>\u305d\u306e\u3088\u3046\u306a\u4e2d\u300cAn Illusion of Progress? Assessing the Current State of Web Agents\u00a0\u300d\u3067\u306f\u300cSurprisingly, many recent agents, except for Operator, do not outperform the simple SeeAct agent (Zheng et al , 2024) released in early 2024.\u300d\u3068\u3082\u6307\u6458\u3055\u308c\u3066\u3044\u308b\u3002\u540c\u8ad6\u6587\u306b\u3082\u3042\u308b\u901a\u308a\u3001\u6b63\u3057\u3044\u8a55\u4fa1\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3084\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u304c\u6c42\u3081\u3089\u308c\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents&nbsp;<\/strong>[30.3]<br>\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u5229\u7528\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u3001\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u3084\u30e2\u30d0\u30a4\u30eb\u30c7\u30d0\u30a4\u30b9\u306e\u30b0\u30e9\u30d5\u30a3\u30ab\u30eb\u30e6\u30fc\u30b6\u30a4\u30f3\u30bf\u30d5\u30a7\u30fc\u30b9(GUI)\u3068\u76f4\u63a5\u5bfe\u8a71\u3059\u308b\u3053\u3068\u3067\u3001\u30c7\u30b8\u30bf\u30eb\u30bf\u30b9\u30af\u3092\u81ea\u52d5\u5316\u3059\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u69d8\u3005\u306a\u30b8\u30a7\u30cd\u30e9\u30ea\u30b9\u30c8\u304a\u3088\u3073\u30b9\u30da\u30b7\u30e3\u30ea\u30b9\u30c8\u30e2\u30c7\u30eb\u306b\u307e\u305f\u304c\u3063\u3066\u8a8d\u77e5\u7684\u8cac\u4efb\u3092\u59d4\u8b72\u3059\u308b\u65b0\u3057\u3044\u69cb\u6210\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3042\u308bAgens S2\u3092\u7d39\u4ecb\u3059\u308b\u3002 Agent S2\u306f\u30013\u3064\u306e\u8457\u540d\u306a\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u4f7f\u7528\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u4e0a\u3067\u306eSOTA(State-of-the-art)\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u78ba\u7acb\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2504.00906v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2504.00906v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Tue, 01 Apr 2025 15:40:27 GMT)<\/li>\n\n\n\n<li><a href=\"https:\/\/devneko.jp\/wordpress\/?p=5611\">Agent S: An Open Agentic Framework that Uses Computers Like a Human&nbsp; \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb<\/a>\u306e\u30d0\u30fc\u30b8\u30e7\u30f32\u3001\u5168\u822c\u7684\u306b\u6027\u80fd\u304c\u4e0a\u304c\u308a\u69d8\u3005\u306a\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067SoTA\u3092\u4e3b\u5f35\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/simular-ai\/Agent-S\">GitHub &#8211; simular-ai\/Agent-S: Agent S: an open agentic framework that uses computers like a human<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>An Illusion of Progress? Assessing the Current State of Web Agents\u00a0<\/strong>[49.8]<br>\u6211\u3005\u306f,Web\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u73fe\u72b6\u3092\u5305\u62ec\u7684\u304b\u3064\u53b3\u5bc6\u306b\u8a55\u4fa1\u3059\u308b\u3002 \u7d50\u679c\u306f\u3001\u73fe\u5728\u306e\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u80fd\u529b\u306e\u975e\u5e38\u306b\u7570\u306a\u308b\u63cf\u5199\u3092\u63cf\u3044\u3066\u304a\u308a\u3001\u4ee5\u524d\u5831\u544a\u3055\u308c\u305f\u7d50\u679c\u306b\u904e\u5ea6\u306b\u6700\u9069\u5316\u3055\u308c\u3066\u3044\u308b\u3053\u3068\u3092\u793a\u5506\u3057\u3066\u3044\u308b\u3002 \u30aa\u30f3\u30e9\u30a4\u30f3\u8a55\u4fa1\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308bOnline-Mind2Web\u3092\u7d39\u4ecb\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2504.01382v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2504.01382v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 02 Apr 2025 05:51:29 GMT)<\/li>\n\n\n\n<li>WEB\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u305f\u3081\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3002\u300cMany recent agents, except for Operator (OpenAI, 2025), underperform the simple SeeAct agent (Zheng et al , 2024) released in early 2024.  Even Operator only achieves a success rate of 61%, showing substantial room for improvement.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/OSU-NLP-Group\/Online-Mind2Web\">GitHub &#8211; OSU-NLP-Group\/Online-Mind2Web<\/a><\/li>\n<\/ul>\n\n\n\n<p>(*1) \u52d5\u304d\u304c\u9762\u767d\u3044\u306e\u3067OpenManus\u3092\u3064\u304b\u3063\u3066\u7121\u7406\u3084\u308a\u5bfe\u5fdc\u3057\u3066\u3044\u308b\u3002\u4eca\u306e\u3068\u3053\u308d\u5b9f\u7528\u6027\u306f\u7591\u554f\u3060\u304c\u3001\u8fd1\u3044\u3046\u3061\u306b\u30d0\u30fc\u30b8\u30e7\u30f3\u30a2\u30c3\u30d7\u4e88\u5b9a\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u4ee5\u524d\u53d6\u308a\u4e0a\u3052\u305fAgent S\u306e\u30d0\u30fc\u30b8\u30e7\u30f32\u304c\u51fa\u3066\u3044\u305f\u3002\u534a\u5e74\u3067OS World\u306e\u30b9\u30b3\u30a2\u304c20.5\u304b\u308927.0\uff0815Step\uff09\u306b\u4e0a\u304c\u3063\u3066\u304a\u308a\u3001\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\uff08LLM\uff09\u306e\u6027\u80fd\u5411\u4e0a\u3082\u3042\u308b\u3060\u308d\u3046\u304c\u7740\u5b9f\u306a\u9032\u5316\u3092\u611f\u3058\u308b\u3002Introduc &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=6530\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Agent S2, Devin 2,\u00a0Amazon Nova Act, An Illusion of Progress? Assessing the Current State of Web Agents\u00a0&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[42,180,181],"class_list":["post-6530","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-autonomous-agent","tag-gui","tag-gui-agent"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6530","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=6530"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6530\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6530"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6530"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6530"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}