{"id":8186,"date":"2026-02-06T05:20:00","date_gmt":"2026-02-05T20:20:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8186"},"modified":"2026-02-01T10:35:22","modified_gmt":"2026-02-01T01:35:22","slug":"causalembed-auto-regressive-multi-vector-generation-in-latent-space-for-visual-document-embedding","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8186","title":{"rendered":"CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding\u00a0<\/strong>[71.9]<br>\u30de\u30eb\u30c1\u30d9\u30af\u30c8\u30eb\u57cb\u3081\u8fbc\u307f\u69cb\u7bc9\u306e\u305f\u3081\u306e\u81ea\u52d5\u56de\u5e30\u751f\u6210\u624b\u6cd5CausalEmbed\u3092\u63d0\u6848\u3059\u308b\u3002 \u30b3\u30f3\u30c8\u30e9\u30b9\u30c8\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u4e2d\u306b\u53cd\u5fa9\u7684\u30de\u30fc\u30b8\u30f3\u640d\u5931\u3092\u5c0e\u5165\u3059\u308b\u3053\u3068\u3067\u3001CausalEmbed\u306f\u57cb\u3081\u8fbc\u307f\u30e2\u30c7\u30eb\u306b\u30b3\u30f3\u30d1\u30af\u30c8\u3067\u3088\u304f\u69cb\u9020\u5316\u3055\u308c\u305f\u8868\u73fe\u3092\u5b66\u3076\u3053\u3068\u3092\u5968\u52b1\u3059\u308b\u3002 \u672c\u624b\u6cd5\u306f,\u6570\u5341\u500b\u306e\u8996\u899a\u30c8\u30fc\u30af\u30f3\u3092\u7528\u3044\u305f\u52b9\u7387\u7684\u306aVDR\u51e6\u7406\u3092\u5b9f\u73fe\u3057,\u30c8\u30fc\u30af\u30f3\u6570\u309230\uff5e15\u500d\u524a\u6e1b\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.21262v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.21262v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 29 Jan 2026 04:47:27 GMT)<\/li>\n\n\n\n<li>\u300cIn this paper, we propose a novel paradigm, CAUSALEMBED, which generates multi-vector embeddings in an auto-regressive manner with significantly shorter sequence lengths. Extensive experiments demonstrate that CAUSALEMBED outperforms pruning-based baselines, achieving superior performance at an extreme 30\u00d7 compression ratio.\u300d\u3068AR\u30e2\u30c7\u30eb\u3092\u6d3b\u7528\u3057\u305f\u52b9\u7387\u7684\u306aEmbedding\u624b\u6cd5\u306e\u63d0\u6848\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":[124,251],"class_list":["post-8186","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-embedding","tag-mllm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8186","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=8186"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8186\/revisions"}],"predecessor-version":[{"id":8187,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8186\/revisions\/8187"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8186"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8186"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8186"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}