{"id":3551,"date":"2023-07-13T05:21:00","date_gmt":"2023-07-12T20:21:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=3551"},"modified":"2023-07-13T05:21:00","modified_gmt":"2023-07-12T20:21:00","slug":"t-mars-text-masking-and-re-scoring","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=3551","title":{"rendered":"T-MARS: Text-Masking and Re-Scoring"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>T-MARS: Improving Visual Representations by Circumventing Text Feature Learning&nbsp;<\/strong>[96.9]<br>LAION\u306e\u753b\u50cf\u306e40%\u8fd1\u304f\u306f\u3001\u5b57\u5e55\u3068\u91cd\u306a\u308b\u30c6\u30ad\u30b9\u30c8\u3092\u542b\u3093\u3067\u3044\u308b\u3068\u3044\u3046\u3001\u6211\u3005\u306e\u89b3\u5bdf\u306b\u52d5\u6a5f\u3065\u3051\u3089\u308c\u305f\u65b0\u3057\u3044\u30c7\u30fc\u30bf\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u624b\u6cd5\u3092\u63d0\u6848\u3059\u308b\u3002 \u6211\u3005\u306e\u30b7\u30f3\u30d7\u30eb\u3067\u30b9\u30b1\u30fc\u30e9\u30d6\u30eb\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u3067\u3042\u308bT-MARS\u306f\u3001\u30c6\u30ad\u30b9\u30c8\u304c\u6b8b\u308a\u306e\u8996\u899a\u7684\u7279\u5fb4\u3092\u652f\u914d\u3059\u308b\u30da\u30a2\u306e\u307f\u3092\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2307.03132v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2307.03132v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Thu, 6 Jul 2023 16:59:52 GMT)<\/li>\n\n\n\n<li>\u52b9\u679c\u7684\u306a\u30c7\u30fc\u30bf\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u624b\u6cd5\u306e\u63d0\u6848\u3001\u753b\u50cf\u304b\u3089\u30c6\u30ad\u30b9\u30c8\u90e8\u5206\u3092\u30de\u30b9\u30ad\u30f3\u30b0\u3001\u30ad\u30e3\u30d7\u30b7\u30e7\u30f3\u3068\u306e\u985e\u4f3c\u5ea6\u3092\u53d6\u308b\u30b7\u30f3\u30d7\u30eb\u306a\u624b\u6cd5\u3002\u300cOur proposed approach is based on the interesting observation that a large fraction of image-caption pairs in web-scale datasets contain images dominated by text features.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/locuslab\/T-MARS\">GitHub &#8211; locuslab\/T-MARS: Code for T-MARS data filtering<\/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":[512,613],"class_list":["post-3551","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-512","tag-613"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3551","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=3551"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3551\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3551"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3551"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3551"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}