{"id":4399,"date":"2024-02-12T05:46:00","date_gmt":"2024-02-11T20:46:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4399"},"modified":"2024-02-12T05:46:00","modified_gmt":"2024-02-11T20:46:00","slug":"mamba%e3%81%aeicl%ef%bc%88in-context-learning%ef%bc%89%e6%80%a7%e8%83%bd","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4399","title":{"rendered":"Mamba\u306eICL\uff08In Context Learning\uff09\u6027\u80fd"},"content":{"rendered":"\n<p>Mamba\u306eICL\u6027\u80fd\u306b\u95a2\u3057\u3066\u8ad6\u6587\u304c\u4e8c\u3064\u51fa\u3066\u3044\u305f\u3002\u7d50\u5c40\u30bf\u30b9\u30af\u306b\u3088\u308b\u3063\u307d\u3044\u3068\u3044\u3046\u611f\u3058\u3060\u308d\u3046\u304b\u3002\u3002\u3002\u5c11\u306a\u304f\u3068\u3082\u4e00\u5b9a\u306eICL\u80fd\u529b\u304c\u3042\u308b\u306e\u306f\u9593\u9055\u3044\u306a\u3044\u3088\u3046\u306b\u601d\u3048\u308b\u3002\u4e00\u3064\u76ee\u306e\u30cf\u30a4\u30d6\u30ea\u30c3\u30c9\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306e\u63d0\u6848\u306f\u3042\u308a\u306a\u306e\u304b\u3001\u305d\u308c\u3060\u3068Mamba\u306e\u826f\u3055\u304c\u8584\u304f\u306a\u308b\u306e\u304b\u60a9\u307e\u3057\u3044\u3068\u3053\u308d\u3067\u306f\u3042\u308b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks\u00a0<\/strong>[26.2]<br>\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb(SSM)\u306f\u8a00\u8a9e\u30e2\u30c7\u30ea\u30f3\u30b0\u306b\u304a\u3051\u308b\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306e\u4ee3\u66ff\u3068\u3057\u3066\u63d0\u6848\u3055\u308c\u3066\u3044\u308b\u3002 \u672c\u7814\u7a76\u3067\u306f,\u5404\u7a2e\u30bf\u30b9\u30af\u3092\u5bfe\u8c61\u3068\u3057\u305f\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30e2\u30c7\u30eb\u306b\u5bfe\u3057\u3066,\u30de\u30f3\u30d0\u306b\u7740\u76ee\u3057\u305fSSM\u306eICL\u6027\u80fd\u3092\u8a55\u4fa1\u3059\u308b\u3002 \u305d\u306e\u7d50\u679c\u3001SSM\u306f\u6a19\u6e96\u56de\u5e30ICL\u30bf\u30b9\u30af\u306b\u304a\u3044\u3066\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u3068\u76f8\u5bb9\u308c\u306a\u3044\u6027\u80fd\u3092\u793a\u3057\u3001\u30b9\u30d1\u30fc\u30b9\u30d1\u30ea\u30c6\u30a3\u5b66\u7fd2\u306e\u3088\u3046\u306a\u30bf\u30b9\u30af\u3067\u306f\u512a\u308c\u3066\u3044\u305f\u3002 \u3053\u308c\u3089\u306e\u5236\u7d04\u306b\u5bfe\u51e6\u3059\u308b\u305f\u3081\u3001\u6211\u3005\u306fMamba\u3068\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30d6\u30ed\u30c3\u30af\u3092\u7d44\u307f\u5408\u308f\u305b\u305f\u30cf\u30a4\u30d6\u30ea\u30c3\u30c9\u30e2\u30c7\u30eb\u3092\u5c0e\u5165\u3057\u3001\u500b\u5225\u306b\u82e6\u52b4\u3059\u308b\u30bf\u30b9\u30af\u306b\u304a\u3044\u3066\u500b\u3005\u306e\u30e2\u30c7\u30eb\u3092\u4e0a\u56de\u308b\u3088\u3046\u306b\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.04248v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.04248v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 6 Feb 2024 18:56:35 GMT)<\/li>\n\n\n\n<li>\u3053\u3061\u3089\u306f\u300cOur results show that SSMs perform comparably to Transformers in standard regression ICL tasks, while outperforming them in tasks like sparse parity learning.However, SSMs fall short in tasks involving non-standard retrieval functionality.\u300d\u3068\u306e\u3053\u3068\u3067\u30bf\u30b9\u30af\u306b\u4f9d\u5b58\u3068\u3044\u3046\u5831\u544a<\/li>\n\n\n\n<li>\u4e0a\u8a18\u3092\u53d7\u3051\u3066MambaFormer\u00a0\u3068\u3044\u3046\u30cf\u30a4\u30d6\u30ea\u30c3\u30c9\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3092\u63d0\u6848<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Is Mamba Capable of In-Context Learning?\u00a0<\/strong>[68.3]<br>Mamba\u306f\u3001\u65b0\u3057\u304f\u63d0\u6848\u3055\u308c\u305f\u9078\u629e\u7684\u306a\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u30de\u30e0\u30d0\u306f\u6587\u8108\u5185\u5b66\u7fd2\u306b\u304a\u3051\u308b\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30e2\u30c7\u30eb\u306e\u6027\u80fd\u3068\u4e00\u81f4\u3059\u308b\u3053\u3068\u3092\u793a\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.03170v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.03170v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 5 Feb 2024 16:39:12 GMT)<\/li>\n\n\n\n<li>\u3053\u3061\u3089\u306f\u300cMamba matches the performance of transformer models for ICL.\u300d\u3068\u306e\u5831\u544a<\/li>\n\n\n\n<li>\u300cMamba appears to solve ICL problems by incrementally refining its internal representations in a manner akin to an iterative optimization strategy, as transformer do.\u300d\u3068\u3044\u3046\u6307\u6458\u3082\u8208\u5473\u6df1\u3044<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Mamba\u306eICL\u6027\u80fd\u306b\u95a2\u3057\u3066\u8ad6\u6587\u304c\u4e8c\u3064\u51fa\u3066\u3044\u305f\u3002\u7d50\u5c40\u30bf\u30b9\u30af\u306b\u3088\u308b\u3063\u307d\u3044\u3068\u3044\u3046\u611f\u3058\u3060\u308d\u3046\u304b\u3002\u3002\u3002\u5c11\u306a\u304f\u3068\u3082\u4e00\u5b9a\u306eICL\u80fd\u529b\u304c\u3042\u308b\u306e\u306f\u9593\u9055\u3044\u306a\u3044\u3088\u3046\u306b\u601d\u3048\u308b\u3002\u4e00\u3064\u76ee\u306e\u30cf\u30a4\u30d6\u30ea\u30c3\u30c9\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306e\u63d0\u6848\u306f\u3042\u308a\u306a\u306e\u304b\u3001\u305d\u308c\u3060\u3068M &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=4399\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Mamba\u306eICL\uff08In Context Learning\uff09\u6027\u80fd&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[200,235],"class_list":["post-4399","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-in-context-learning","tag-mamba"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4399","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=4399"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4399\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4399"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4399"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4399"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}