{"id":1379,"date":"2025-08-27T20:39:00","date_gmt":"2025-08-27T12:39:00","guid":{"rendered":"http:\/\/www.applicationsofllm.com\/?p=1379"},"modified":"2025-09-17T20:43:42","modified_gmt":"2025-09-17T12:43:42","slug":"%e9%98%bf%e9%87%8c%e5%bc%80%e6%ba%90%e9%80%9a%e4%b9%89deepresearch%ef%bc%8c%e6%80%a7%e8%83%bd%e8%b6%85openai%e3%80%81deepseek%e6%97%97%e8%88%b0%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"http:\/\/www.applicationsofllm.com\/index.php\/2025\/08\/27\/%e9%98%bf%e9%87%8c%e5%bc%80%e6%ba%90%e9%80%9a%e4%b9%89deepresearch%ef%bc%8c%e6%80%a7%e8%83%bd%e8%b6%85openai%e3%80%81deepseek%e6%97%97%e8%88%b0%e6%a8%a1%e5%9e%8b\/","title":{"rendered":"\u963f\u91cc\u5f00\u6e90\u901a\u4e49DeepResearch\uff0c\u6027\u80fd\u8d85OpenAI\u3001DeepSeek\u65d7\u8230\u6a21\u578b"},"content":{"rendered":"\n

9\u670817\u65e5\u51cc\u6668\uff0c\u963f\u91cc\u5f00\u6e90\u65d7\u4e0b\u9996\u4e2a\u6df1\u5ea6\u7814\u7a76Agent\u6a21\u578b\u2014\u2014\u901a\u4e49DeepResearch\u3002\u8be5\u6a21\u578b\u5728HLE\u3001BrowseComp-zh\u3001GAIA\u7b49\u591a\u4e2a\u6743\u5a01\u8bc4\u6d4b\u96c6\u4e0a\u53d6\u5f97SOTA\u6210\u7ee9(State-of-the-art)\uff0c\u8d85\u8d8aOpenAI Deep Research\u3001DeepSeek-V3.1\u7b49Agent\u6a21\u578b\u3002\u76ee\u524d\uff0c\u901a\u4e49DeepResearch\u7684\u6a21\u578b\u3001\u6846\u67b6\u548c\u65b9\u6848\u5747\u5df2\u5168\u9762\u5f00\u6e90\uff0c\u7528\u6237\u53ef\u5728Github\u3001Hugging Face\u548c\u9b54\u642d\u793e\u533a\u793e\u533a\u4e0b\u8f7d\u6a21\u578b\u548c\u4ee3\u7801\u3002<\/p>\n\n\n\n

\u901a\u4e49DeepResearch\u767b\u9876\u5f00\u6e90\u7b2c\u4e00<\/p>\n\n\n\n

\u6df1\u5ea6\u7814\u7a76\u662f\u8fd1\u4e00\u5e74\u6765\u7684\u7814\u7a76\u70ed\u70b9\uff0c\u8c37\u6b4c\u3001OpenAI\u3001Anthropic\u7b49\u4e3b\u6d41AI\u4f01\u4e1a\u5747\u5728\u6295\u5165\u76f8\u5173\u6280\u672f\u548c\u4ea7\u54c1\u3002\u7136\u800c\uff0c\u73b0\u6709\u65b9\u6cd5\u666e\u904d\u91c7\u7528\u201c\u5355\u7a97\u53e3\u3001\u7ebf\u6027\u7d2f\u52a0\u201d\u7684\u4fe1\u606f\u5904\u7406\u6a21\u5f0f\uff0c\u6240\u6709\u4e2d\u95f4\u601d\u8def\u548c\u68c0\u7d22\u5230\u7684\u4fe1\u606f\u5806\u79ef\u5728\u5355\u4e00\u4e0a\u4e0b\u6587\u4e2d\u3002\u8fd9\u79cd\u6a21\u5f0f\u5728\u5904\u7406\u957f\u5468\u671f\u4efb\u52a1\u65f6\uff0cAgent\u4f1a\u9762\u4e34\u201c\u8ba4\u77e5\u7a7a\u95f4\u7a92\u606f\u201d\u548c\u201c\u4e0d\u53ef\u9006\u7684\u566a\u58f0\u6c61\u67d3\u201d\u7684\u6311\u6218\uff0c\u5bfc\u81f4\u63a8\u7406\u80fd\u529b\u4e0b\u964d\uff0c\u6700\u7ec8\u96be\u4ee5\u5b8c\u6210\u771f\u6b63\u957f\u7a0b\u3001\u590d\u6742\u7684\u7814\u7a76\u4efb\u52a1\u3002<\/p>\n\n\n\n

\u4e3a\u6b64\uff0c\u901a\u4e49\u56e2\u961f\u6784\u5efa\u4e86\u4e00\u5957\u4ee5\u5408\u6210\u6570\u636e\u9a71\u52a8\u3001\u8d2f\u7a7f\u9884\u8bad\u7ec3\u4e0e\u540e\u8bad\u7ec3\u7684\u5b8c\u6574\u8bad\u7ec3\u94fe\u8def\u3002\u8be5\u94fe\u8def\u4ee5Qwen3-30B-A3B \u6a21\u578b\u4e3a\u57fa\u5ea7\u8fdb\u884c\u4f18\u5316\uff0c\u4e3a\u4e86\u786e\u4fdd\u8bad\u7ec3\u7684\u9ad8\u6548\u6027\u4e0e\u6a21\u578b\u7684\u9c81\u68d2\u6027\uff0c\u56e2\u961f\u521b\u65b0\u6027\u5730\u8bbe\u8ba1\u4e86\u8986\u76d6\u771f\u5b9e\u73af\u5883\u4e0e\u865a\u62df\u73af\u5883\u7684RL\u7b97\u6cd5\u9a8c\u8bc1\u4e0e\u771f\u5b9e\u8bad\u7ec3\u6a21\u5757\uff0c\u5e76\u7ed3\u5408\u9ad8\u6548\u5f02\u6b65\u5f3a\u5316\u5b66\u4e60\u7b97\u6cd5\u53ca\u81ea\u52a8\u5316\u6570\u636e\u7b56\u5c55\uff08Data Curation\uff09\u6d41\u7a0b\uff0c\u663e\u8457\u63d0\u5347\u4e86\u6a21\u578b\u7684\u8fed\u4ee3\u901f\u5ea6\u548c\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n\n\n\n

\u5728\u63a8\u7406\u9636\u6bb5\uff0c\u56e2\u961f\u8bbe\u8ba1\u4e86 ReAct \u548c\u57fa\u4e8e\u81ea\u7814\u7684IterResearch\u7684Heavy \u4e24\u79cd\u6a21\u5f0f\uff0c\u524d\u8005\u7528\u4e8e\u7cbe\u51c6\u8003\u5bdf\u6a21\u578b\u7684\u57fa\u7840\u5185\u5728\u80fd\u529b\uff0c\u540e\u8005\u5219\u901a\u8fc7 test-time scaling \u7b56\u7565\uff0c\u5145\u5206\u6316\u6398\u5e76\u5c55\u73b0\u4e86\u6a21\u578b\u6240\u80fd\u8fbe\u5230\u7684\u6027\u80fd\u4e0a\u9650\uff0c\u5373\u4f7f\u5728\u957f\u4efb\u52a1\u4e2d\uff0c\u4e5f\u80fd\u5b9e\u73b0\u9ad8\u8d28\u91cf\u7684\u63a8\u7406\u3002<\/p>\n\n\n\n

\"\"\/<\/figure>\n\n\n\n

\u5728Humanity’s Last Exam\uff08HLE\uff09\u3001BrowseComp\u3001BrowseComp-ZH\u3001GAIA\u3001xbench-deepsearch\u3001WebWalkerQA\u4ee5\u53caFrames\u7b49\u6743\u5a01Agent\u8bc4\u6d4b\u96c6\u4e0a\uff0c\u901a\u4e49DeepResearch\u6a21\u578b\u4ee53B\u6fc0\u6d3b\u53c2\u6570\uff0c\u6027\u80fd\u8d85\u8d8a\u57fa\u4e8eOpenAI o3\u3001DeepSeek V3.1\u548cClaude-4-Sonnet\u7b49\u65d7\u8230\u6a21\u578b\u7684ReAct Agent\u3002<\/p>\n\n\n\n

\u4eca\u5e74\u4ee5\u6765\uff0c\u963f\u91cc\u5df2\u8fde\u7eed\u5f00\u6e90WebWalker\u3001WebDancer\u548cWebSailor\u7b49\u591a\u6b3e\u68c0\u7d22\u548c\u63a8\u7406\u667a\u80fd\u4f53\uff0c\u5e76\u5168\u90e8\u65a9\u83b7\u5f00\u6e90SOTA\u6210\u7ee9\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

9\u670817\u65e5\u51cc\u6668\uff0c\u963f\u91cc\u5f00\u6e90\u65d7\u4e0b\u9996\u4e2a\u6df1\u5ea6\u7814\u7a76Agent\u6a21\u578b\u2014\u2014\u901a\u4e49DeepResearch\u3002\u8be5\u6a21\u578b\u5728HLE\u3001Bro […]<\/p>\n","protected":false},"author":1,"featured_media":1380,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[],"class_list":["post-1379","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-8"],"_links":{"self":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1379","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/comments?post=1379"}],"version-history":[{"count":1,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1379\/revisions"}],"predecessor-version":[{"id":1381,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1379\/revisions\/1381"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media\/1380"}],"wp:attachment":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media?parent=1379"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/categories?post=1379"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/tags?post=1379"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}