{"id":7979,"date":"2026-01-02T05:00:00","date_gmt":"2026-01-01T20:00:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7979"},"modified":"2025-12-27T14:02:40","modified_gmt":"2025-12-27T05:02:40","slug":"geneval-2-addressing-benchmark-drift-in-text-to-image-evaluation","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7979","title":{"rendered":"GenEval 2: Addressing Benchmark Drift in Text-to-Image Evaluation"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>GenEval 2: Addressing Benchmark Drift in Text-to-Image Evaluation\u00a0<\/strong>[115.5]<br>\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u30c9\u30ea\u30d5\u30c8\u306f\u3001\u6700\u3082\u4eba\u6c17\u306e\u3042\u308bT2I\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u4e00\u3064\u3067\u3042\u308bGenEval\u306b\u3068\u3063\u3066\u91cd\u8981\u306a\u554f\u984c\u3067\u3042\u308b\u3053\u3068\u3092\u793a\u3059\u3002 \u6211\u3005\u306f\u65b0\u3057\u3044\u30d9\u30f3\u30c1\u30de\u30fc\u30afGenEval 2\u3092\u5c0e\u5165\u3057\u3001\u539f\u59cb\u7684\u306a\u8996\u899a\u6982\u5ff5\u306e\u30ab\u30d0\u30ec\u30c3\u30b8\u3092\u6539\u5584\u3057\u3001\u3088\u308a\u9ad8\u5ea6\u306a\u69cb\u6210\u6027\u3092\u5b9f\u73fe\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2512.16853v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2512.16853v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 18 Dec 2025 18:26:56 GMT)<\/li>\n\n\n\n<li>\u300cGenEval was well-aligned with human judgment at the time of its release, it has drifted far from human judgment over time\u2014resulting in an absolute error of as much as 17.7% for current models. This level of drift strongly suggests that GenEval has been saturated for some time, as we verify via a large-scale human study. To help fill this benchmarking gap, we introduce a new benchmark, GenEval 2, with improved coverage of primitive visual concepts and higher degrees of compositionality, which we show is more challenging for current models.\u300d\u3068GenEval\u306e\u65b0\u305f\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u306e\u63d0\u6848\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/facebookresearch\/GenEval2\">GitHub &#8211; facebookresearch\/GenEval2: Evaluation codes and data for GenEval2<\/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":[517],"class_list":["post-7979","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-517"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7979","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=7979"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7979\/revisions"}],"predecessor-version":[{"id":7980,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7979\/revisions\/7980"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7979"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7979"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7979"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}