{"id":10408,"date":"2026-07-26T06:58:59","date_gmt":"2026-07-25T22:58:59","guid":{"rendered":"https:\/\/infernews.com\/blog\/activevision\/"},"modified":"2026-07-26T07:01:56","modified_gmt":"2026-07-25T23:01:56","slug":"activevision","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/activevision\/","title":{"rendered":"ActiveVision \u9ede\u51fa\u8996\u89ba\u63a8\u7406\u771f\u7a7a\u5e36"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/banner-2.jpg\" alt=\"ActiveVision \u2014 An Exam for Active Observers. Vision is a loop, not a glance.\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u4e0d\u5c11\u8996\u89ba\u984c\u76ee\u5514\u4fc2\u9760\u4e00\u773c\u8fa8\u8a8d\uff0c\u800c\u4fc2\u8981\u6cbf\u4f4f\u7dda\u8ffd\u3001\u9010\u5340\u57df\u6578\u3001\u4e00\u6b65\u6b65\u6838\u5c0d\u5148\u7b54\u5f97\u5230\uff1bActiveVision \u6b63\u6b63\u91dd\u5c0d\u5462\u7a2e\u843d\u5dee\u800c\u4f86\u3002\u4f5c\u70ba\u4e00\u500b benchmark\uff0c\u5b83\u96c6\u4e2d\u6e2c\u8a66 iterative visual reasoning\uff0c\u8655\u7406\u7684\u662f\u6a21\u578b\u770b\u5f97\u5230\u756b\u9762\uff0c\u4f46\u672a\u5fc5\u80fd\u6301\u7e8c\u6574\u7406\u89c0\u5bdf\u904e\u7a0b\u7684\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6709\u591a\u6a21\u614b\u6a21\u578b\u5e38\u898b\u505a\u6cd5\u662f\u5c0d\u55ae\u5f35\u5716\u4f5c\u4e00\u6b21\u6027\u5224\u8b80\uff0c\u518d\u914d\u5408 chain-of-thought \u76f4\u63a5\u4f5c\u7b54\uff1b\u4f5c\u8005\u8a8d\u70ba\u9019\u7a2e single-glance \u7bc4\u5f0f\uff0c\u5c0d\u9700\u8981\u53cd\u8986\u6383\u63cf\u3001\u8ffd\u8e64\u9806\u5e8f\u8207\u7dad\u6301\u4e2d\u9593\u72c0\u614b\u7684\u984c\u578b\u7279\u5225\u5403\u529b\u3002ActiveVision \u56e0\u6b64\u8a2d\u8a08\u4e86 17 \u500b\u4efb\u52d9\uff0c\u4e26\u7528 deterministic program \u751f\u6210\u5834\u666f\uff0c\u518d\u4ee5 photorealistic \u65b9\u5f0f\u91cd\u7e6a\uff0c\u4ee4\u756b\u9762\u81ea\u7136\u4e4b\u9918\u4ecd\u4fdd\u7559\u53ef\u9a57\u8b49\u7d50\u69cb\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img data-dominant-color=\"f1ede4\" data-has-transparency=\"false\" style=\"--dominant-color: #f1ede4;\" loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"686\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/image-3.png\" alt=\"\" class=\"wp-image-10410 not-transparent\" srcset=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/image-3.png 1024w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/image-3-300x201.png 300w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/image-3-768x515.png 768w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u6578\u5b57\u76f8\u7576\u76f4\u63a5\uff1a\u4eba\u985e\u8868\u73fe\u70ba 96.1%\uff0c\u524d\u6cbf\u6a21\u578b\u5728\u5b98\u65b9\u7121\u5de5\u5177\u8a55\u6e2c\u4e0b\u6700\u9ad8\u7d04 10.6%\uff0c\u5dee\u8ddd\u63a5\u8fd1 9 \u500d\u3002\u7db2\u7ad9\u4ea6\u5217\u51fa agent \u7248\u672c\u7684 tool-use ablation\uff0c\u50cf Claude Code \u8207 Codex \u63a5\u5165\u5de5\u5177\u5f8c\uff0c\u5206\u6578\u660e\u986f\u9ad8\u904e\u7d14 chain-of-thought\uff0c\u8868\u793a\u554f\u984c\u672a\u5fc5\u53ea\u662f\u300c\u770b\u4e0d\u61c2\u5716\u300d\uff0c\u800c\u662f\u7f3a\u5c11\u53ef\u9010\u6b65\u5916\u5316\u8207\u64cd\u4f5c\u7684\u89e3\u984c\u6d41\u7a0b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6536\u9304 17 \u500b\u4efb\u52d9\uff0c\u91cd\u9ede\u653e\u5728 distributed scanning \u8207 sequential traversal \u4e00\u985e\u9010\u6b65\u89c0\u5bdf\u984c<\/li>\n\n\n\n<li>\u5b98\u65b9\u8a55\u6e2c\u6db5\u84cb Claude\u3001GPT\u3001Gemini\uff0c\u4ea6\u63d0\u4f9b agent ablation \u8173\u672c<\/li>\n\n\n\n<li>\u6578\u64da\u96c6\u53ef\u7d93 Hugging Face \u4e0b\u8f09\uff0c\u8a55\u6e2c\u7a0b\u5f0f\u4ee5 Python \u70ba\u4e3b<\/li>\n\n\n\n<li>\u540c\u4e00\u975c\u614b\u5716\u7247\u4e5f\u80fd\u8feb\u4f7f\u6a21\u578b\u505a\u591a\u6b65\u63a8\u7406\uff0c\u5514\u9760\u5f71\u7247\u8f38\u5165\u6490\u8d77\u96e3\u5ea6<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u500b GitHub \u9805\u76ee\u6bd4\u8f03\u50cf\u7814\u7a76\u8207\u8a55\u6e2c\u57fa\u5efa\uff0c\u800c\u5514\u4fc2\u5373\u7528\u578b\u7522\u54c1\uff1a\u4f60\u9700\u8981\u5148\u4e0b\u8f09\u6578\u64da\u96c6\u3001\u914d\u7f6e\u5c0d\u61c9\u4f9b\u61c9\u5546 API\uff0c\u7136\u5f8c\u7528 repo \u5167\u7684 eval \u8173\u672c\u8dd1\u7d50\u679c\u3002\u5c0d\u505a\u591a\u6a21\u614b\u6a21\u578b\u8a55\u6e2c\u3001Agentic \u5de5\u4f5c\u6d41\u3001\u6216\u8005\u60f3\u9a57\u8b49 Computer-use agents\u3001CUAs \u5f0f\u5916\u90e8\u5de5\u5177\u5354\u4f5c\u50f9\u503c\u7684\u5718\u968a\uff0c\u5b83\u63d0\u4f9b\u4e86\u4e00\u500b\u5f88\u5c16\u92b3\u7684\u6aa2\u67e5\u9ede\uff1a\u6a21\u578b\u662f\u5426\u771f\u7684\u6703\u300c\u89c0\u5bdf\u300d\uff0c\u9084\u662f\u53ea\u6703\u5c0d\u5f71\u50cf\u4f5c\u9ad8\u968e\u731c\u6e2c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/activevision.dev\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/saccharomycetes\/ActiveVision\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/arxiv.org\/pdf\/2607.16165\" rel=\"noopener noreferrer\"><strong>Paper<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u671b\u4e00\u773c\u7b54\u984c\u7684\u591a\u6a21\u614b\u6a21\u578b\uff0c\u5728 ActiveVision \u5e7e\u4e4e\u96c6\u9ad4\u5931\u9748\u3002\u9019\u500b benchmark \u628a\u300c\u8981\u4e00\u8def\u89c0\u5bdf\u5148\u7b54\u5230\u300d\u7684\u80fd\u529b\uff0c\u62c6\u6210\u66f4\u8cbc\u8fd1\u4eba\u985e\u89e3\u984c\u7bc0\u594f\u7684\u6e2c\u8a66\u3002<\/p>\n","protected":false},"author":8,"featured_media":10407,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,140,153,116,137,152,119,191,199],"tags":[],"class_list":["post-10408","post","type-post","status-publish","format-standard","hentry","category-133","category-gemini","category-openai","category-agentic","category-api","category-python","category-119","category-anthropic","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10408","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=10408"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10408\/revisions"}],"predecessor-version":[{"id":10411,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10408\/revisions\/10411"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10407"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10408"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10408"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10408"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}