{"id":7767,"date":"2025-11-17T05:30:00","date_gmt":"2025-11-16T20:30:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7767"},"modified":"2025-11-15T20:26:26","modified_gmt":"2025-11-15T11:26:26","slug":"tabpfn-2-5-advancing-the-state-of-the-art-in-tabular-foundation-models","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7767","title":{"rendered":"TabPFN-2.5: Advancing the State of the Art in Tabular Foundation Models\u00a0\/ Does TabPFN Understand Causal Structures? \/ TransactionGPT\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>TabPFN-2.5: Advancing the State of the Art in Tabular Foundation Models&nbsp;<\/strong>[76.5]<br>TabPFN-2.5\u306f5\u4e07\u306e\u30c7\u30fc\u30bf\u30dd\u30a4\u30f3\u30c8\u30682,000\u306e\u6a5f\u80fd\u3092\u6301\u3064\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u7528\u306b\u69cb\u7bc9\u3055\u308c\u3066\u3044\u308b\u3002 \u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3055\u308c\u305f\u30c4\u30ea\u30fc\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\u3068AutoGluon 1.4\u306e\u7cbe\u5ea6\u3092\u5927\u5e45\u306b\u4e0a\u56de\u3063\u305f\u3002 \u751f\u7523\u7528\u3068\u3057\u3066,TabPFN-2.5\u3092\u5c0f\u578b\u307e\u305f\u306f\u6728\u88fd\u30a2\u30f3\u30b5\u30f3\u30d6\u30eb\u306b\u5909\u63db\u3059\u308b\u65b0\u3057\u3044\u84b8\u7559\u30a8\u30f3\u30b8\u30f3\u3092\u5c0e\u5165\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2511.08667v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2511.08667v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Thu, 13 Nov 2025 01:01:46 GMT)<\/li>\n\n\n\n<li>\u30c6\u30fc\u30d6\u30eb\u30c7\u30fc\u30bf\u306b\u5bfe\u3059\u308b\u57fa\u76e4\u30e2\u30c7\u30eb\u306e\u63d0\u6848\u3001<a href=\"https:\/\/huggingface.co\/spaces\/TabArena\/leaderboard\">TabArena &#8211; a Hugging Face Space by TabArena<\/a>\u3067\u300cTabPFN-2.5 is now the leading method for the industry standard benchmark TabArena (which contains datasets with up to 100,000 training data points), substantially outperforming tuned tree-based models and matching the accuracy of AutoGluon 1.4, a complex four-hour tuned ensemble that even includes the previous TabPFNv2. Remarkably, default TabPFN-2.5 has a 100% win rate against default XGBoost on small to medium-sized classification datasets (\u226410,000 data points, 500 features) and a 87% win rate on larger datasets up to 100K samples and 2K features (85% for regression).\u300d\u3068\u9ad8\u6027\u80fd\u3092\u4e3b\u5f35<\/li>\n\n\n\n<li><a href=\"https:\/\/priorlabs.ai\/\">Prior Labs<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Does TabPFN Understand Causal Structures?&nbsp;<\/strong>[40.2]<br>\u672c\u7814\u7a76\u3067\u306f,TabPFN\u304c\u5185\u90e8\u8868\u73fe\u306b\u56e0\u679c\u60c5\u5831\u3092\u30a8\u30f3\u30b3\u30fc\u30c9\u3059\u308b\u304b\u3069\u3046\u304b\u3092\u691c\u8a0e\u3059\u308b\u3002 \u5b66\u7fd2\u53ef\u80fd\u306a\u30c7\u30b3\u30fc\u30c0\u3068\u56e0\u679c\u30c8\u30fc\u30af\u30f3\u3092\u7528\u3044\u305f\u30a2\u30c0\u30d7\u30bf\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u958b\u767a\u3057\u305f\u3002 \u8a55\u4fa1\u306e\u7d50\u679c,TabPFN\u306e\u57cb\u3081\u8fbc\u307f\u306b\u306f\u56e0\u679c\u60c5\u5831\u304c\u542b\u307e\u308c\u3066\u304a\u308a,\u5f93\u6765\u306e\u56e0\u679c\u767a\u898b\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3088\u308a\u3082\u512a\u308c\u3066\u3044\u308b\u3053\u3068\u304c\u308f\u304b\u3063\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2511.07236v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2511.07236v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 10 Nov 2025 15:53:15 GMT)<\/li>\n\n\n\n<li>\u300cWe show that TabPFN\u2019s embeddings contain causal information and that our adaptor framework outperforms traditional causal discovery algorithms when causal information is extracted from mid- range layers. This further promotes leveraging pre-trained tabular models for extracting causal structures, improving the interpretability of these models, and aiding in scientific discovery.\u300d\u3068\u8208\u5473\u6df1\u3044\u6027\u8cea\u3092\u5831\u544a\u3002<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>TransactionGPT\u00a0<\/strong>[41.9]<br>TransactionGPT\u306f\u3001\u4e16\u754c\u6700\u5927\u306e\u6c7a\u6e08\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u5185\u306e\u30b3\u30f3\u30b7\u30e5\u30fc\u30de\u30c8\u30e9\u30f3\u30b6\u30af\u30b7\u30e7\u30f3\u30c7\u30fc\u30bf\u306e\u57fa\u76e4\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u652f\u6255\u3044\u30c8\u30e9\u30f3\u30b6\u30af\u30b7\u30e7\u30f3\u30c7\u30fc\u30bf\u306e\u8907\u96d1\u306a\u30c0\u30a4\u30ca\u30df\u30af\u30b9\u3092\u6349\u3048\u308b\u305f\u3081\u306b,\u65b0\u3057\u30443D-Transformer\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3092\u63d0\u6848\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2511.08939v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2511.08939v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 13 Nov 2025 01:20:09 GMT)<\/li>\n\n\n\n<li>Visa Research\u306b\u3088\u308b\u57fa\u76e4\u30e2\u30c7\u30eb\u3002\u300cTransactionGPT (TGPT), a foundation model that captures complex consumer shopping dynamics from Multi-Modal-Temporal-Tabular (MMTT) data.\u300d\u3001\u300cExtensive experiments on large-scale, real-world payment data validate TGPT\u2019s ability to learn meaningful transaction patterns, leading to significant performance improve- ments on critical downstream tasks. Furthermore, we quantify the benefits of several designs that enhance the TGPT\u2019s efficiency and scalability.\u300d\u3068\u306e\u3053\u3068\u3002<\/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":[153,393,498,564],"class_list":["post-7767","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-foundation-models","tag-table","tag-498","tag-564"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7767","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=7767"}],"version-history":[{"count":3,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7767\/revisions"}],"predecessor-version":[{"id":7787,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7767\/revisions\/7787"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7767"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7767"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7767"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}