{"id":7024,"date":"2025-06-30T03:49:00","date_gmt":"2025-06-29T18:49:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7024"},"modified":"2025-06-29T10:34:43","modified_gmt":"2025-06-29T01:34:43","slug":"routing-mamba-memba","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7024","title":{"rendered":"Routing Mamba, Memba"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Routing Mamba: Scaling State Space Models with Mixture-of-Experts Projection\u00a0<\/strong>[88.5]<br>\u7dda\u5f62\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb(SSM)\u306f\u3001\u30b7\u30fc\u30b1\u30f3\u30b9\u30e2\u30c7\u30ea\u30f3\u30b0\u306b\u304a\u3044\u3066\u9855\u8457\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u5411\u4e0a\u3092\u63d0\u4f9b\u3059\u308b\u3002 Mamba\u306e\u3088\u3046\u306a\u6700\u8fd1\u306e\u9032\u6b69\u306f\u3001\u5165\u529b\u4f9d\u5b58\u306e\u30b2\u30fc\u30c6\u30a3\u30f3\u30b0\u3068\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u5bfe\u5fdc\u306e\u5b9f\u88c5\u306b\u3088\u308a\u3001SSM\u3092\u3055\u3089\u306b\u5f37\u5316\u3057\u3066\u3044\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u7dda\u5f62\u5c04\u5f71\u306e\u5c02\u9580\u5bb6\u306b\u3088\u308b\u758e\u6df7\u5408\u3092\u7528\u3044\u3066SSM\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u30b9\u30b1\u30fc\u30eb\u3059\u308b\u65b0\u3057\u3044\u624b\u6cd5\u3067\u3042\u308bRouting Mamba (RoM)\u3092\u7d39\u4ecb\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2506.18145v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2506.18145v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Sun, 22 Jun 2025 19:26:55 GMT)<\/li>\n\n\n\n<li>\u300cWe introduce Routing Mamba (RoM), a novel framework that integrates MoE mechanisms into SSMs by leveraging Mamba\u2019s projection layers as scalable expert components.\u300d\u3068MoE\u7684\u306a\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092Mamba\u306b\u6301\u3061\u8fbc\u3093\u3060\u7814\u7a76\u3002\u52b9\u7387\u30fb\u6027\u80fd\u304c\u5411\u4e0a\u3068\u306e\u3053\u3068\u3002<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Memba: Membrane-driven Parameter-Efficient Fine-Tuning for Mamba\u00a0<\/strong>[21.5]<br>Memba\u306f\u3001State Space Models\uff08SSMs\uff09\u306b\u7279\u5316\u3057\u305f\u65b0\u3057\u3044\u30d1\u30e9\u30e1\u30fc\u30bf\u52b9\u7387\u7684\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u624b\u6cd5\u3067\u3042\u308a\u3001Mamba\u30e2\u30c7\u30eb\u306e\u80fd\u529b\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u3092\u76ee\u6307\u3057\u3066\u3044\u307e\u3059\u3002Leaky Integrate Membrane (LIM)\u30cb\u30e5\u30fc\u30ed\u30f3\u3092\u7528\u3044\u3066\u6642\u9593\u7684\u306a\u60c5\u5831\u4fdd\u6301\u3092\u5f37\u5316\u3057\u3001\u5f93\u6765\u306e\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u624b\u6cd5\u3068\u6bd4\u3079\u3066\u512a\u308c\u305f\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5b9f\u73fe\u3057\u3066\u3044\u307e\u3059\u3002\u5b9f\u9a13\u7d50\u679c\u306f\u3001Memba\u304c\u8a00\u8a9e\u30e2\u30c7\u30eb\u3084\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u30d3\u30b8\u30e7\u30f3\u306e\u30bf\u30b9\u30af\u306b\u304a\u3044\u3066\u4ed6\u306e\u624b\u6cd5\u3088\u308a\u3082\u8457\u3057\u3044\u6539\u5584\u3092\u793a\u3059\u3053\u3068\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2506.18184v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2506.18184v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Sun, 22 Jun 2025 21:52:45 GMT)<\/li>\n\n\n\n<li>Mamba\u7528\u306b\u8a2d\u8a08\u3055\u308c\u305f\u52b9\u7387\u7684\u306afine tuning\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Intelligent-Computing-Lab-Yale\/Memba\">https:\/\/github.com\/Intelligent-Computing-Lab-Yale\/Memba<\/a>\u3068\u306e\u3053\u3068\u3060\u304c\u3001\u73fe\u6642\u70b9\u3067\u306f404<\/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":[235,376],"class_list":["post-7024","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-mamba","tag-ssm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7024","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=7024"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7024\/revisions"}],"predecessor-version":[{"id":7025,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7024\/revisions\/7025"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7024"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7024"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7024"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}