{"id":7975,"date":"2025-12-29T05:51:00","date_gmt":"2025-12-28T20:51:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7975"},"modified":"2025-12-27T13:58:01","modified_gmt":"2025-12-27T04:58:01","slug":"probing-scientific-general-intelligence-of-llms-with-scientist-aligned-workflows","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7975","title":{"rendered":"Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows\u00a0<\/strong>[203.4]<br>PIM(Practical Inquiry Model)\u306b\u57fa\u3065\u304f\u904b\u7528SGI\u5b9a\u7fa9\u3092\u63d0\u6848\u3059\u308b\u3002 \u6df1\u5c64\u7814\u7a76\u3001\u30a2\u30a4\u30c7\u30a2\u751f\u6210\u3001\u30c9\u30e9\u30a4\/\u30a6\u30a7\u30c3\u30c8\u5b9f\u9a13\u3001\u5b9f\u9a13\u63a8\u8ad6\u306e4\u3064\u306e\u30bf\u30b9\u30af\u3092\u901a\u3058\u3066\u904b\u7528\u3057\u3066\u3044\u307e\u3059\u3002 \u79c1\u305f\u3061\u306ePIM\u306b\u3088\u308b\u5b9a\u7fa9\u3001\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u4e2d\u5fc3\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3001\u5b9f\u8a3c\u7684\u306a\u6d1e\u5bdf\u306f\u3001\u771f\u306b\u79d1\u5b66\u7684\u306a\u767a\u898b\u306b\u53c2\u52a0\u3059\u308bAI\u30b7\u30b9\u30c6\u30e0\u306e\u57fa\u76e4\u3092\u78ba\u7acb\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2512.16969v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2512.16969v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 18 Dec 2025 12:44:36 GMT)<\/li>\n\n\n\n<li>scientific general intelligence (SGI)\u3001\u300cSGI is an AI that can autonomously navigate the complete, iterative cycle of scientific inquiry with the versatility and proficiency of a human scientist\u300d\u306e\u7814\u7a76\u3001\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u7b49\u3082\u63d0\u6848\u3057\u3066\u3044\u308b\u3002\u300cExperiments reveal a consistent pattern: in Deep Research, models show step-level alignment but low exact-match accuracy (10\u201320%), with brittleness in quantitative reasoning; in Idea Generation, hypotheses are fluent but underspecified and infeasible; in Dry Experiment, code is executable but PassAll@k remains low; in Wet Experiment, sequences show omissions and misordering; and in Experimental Reasoning, causal reasoning outperforms comparative, with persistent multimodal challenges. These highlight gaps between linguistic fluency and integrated scientific cognition.\u300d\u3068\u3042\u308b\u306a\u3069\u9053\u534a\u3070\u3068\u3044\u3046\u611f\u3058\u3067\u306f\u3042\u308b\u304c\u975e\u5e38\u306b\u6d41\u884c\u3063\u3066\u3044\u308b\u5206\u91ce\u3060\u3068\u601d\u3046\u3002<\/li>\n\n\n\n<li>SGI-Bench\u306e\u4e0a\u4f4d\u306fGemini 3 Pro, Claude Sonnet 4.5, Qwen3 Max, GPT-4.1, GPT-5.2 Pro\u3068\u5404\u793e\u306e\u30d5\u30ed\u30f3\u30c6\u30a3\u30a2\u30e2\u30c7\u30eb\u304c\u4e26\u3076\u3002 <\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/internscience.github.io\/SGI-Page\/\">SGI-Bench \u2014 Scientific General Intelligence<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[623],"class_list":["post-7975","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-623"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7975","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=7975"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7975\/revisions"}],"predecessor-version":[{"id":7976,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7975\/revisions\/7976"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7975"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7975"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7975"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}