{"id":8150,"date":"2026-01-26T04:14:00","date_gmt":"2026-01-25T19:14:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8150"},"modified":"2026-01-26T08:23:36","modified_gmt":"2026-01-25T23:23:36","slug":"understanding-multilingualism-in-mixture-of-experts-llms-routing-mechanism-expert-specialization-and-layerwise-steering","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8150","title":{"rendered":"Understanding Multilingualism in Mixture-of-Experts LLMs: Routing Mechanism, Expert Specialization, and Layerwise Steering"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Understanding Multilingualism in Mixture-of-Experts LLMs: Routing Mechanism, Expert Specialization, and Layerwise Steering&nbsp;<\/strong>[61.1]<br>\u672c\u7814\u7a76\u3067\u306f,\u4e2d\u9593\u5c64\u306b\u304a\u3051\u308b\u30eb\u30fc\u30c6\u30a3\u30f3\u30b0\u52d5\u4f5c\u3092,\u652f\u914d\u8a00\u8a9e\u306b\u95a2\u9023\u3059\u308b\u5171\u6709\u5c02\u9580\u5bb6\u306b\u9069\u5fdc\u7684\u306b\u8a98\u5c0e\u3059\u308b\u30eb\u30fc\u30c6\u30a3\u30f3\u30b0\u8a98\u5c0e\u578b\u30b9\u30c6\u30a2\u30ea\u30f3\u30b0\u624b\u6cd5\u3092\u63d0\u6848\u3059\u308b\u3002 <br><a href=\"http:\/\/arxiv.org\/abs\/2601.14050v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.14050v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Tue, 20 Jan 2026 15:04:25 GMT)<\/li>\n\n\n\n<li>\u300cLanguages within the same linguistic family tend to share similar routing distributions, whereas linguistically distant languages are routed through more distinct subsets of experts (cf. Section 4.2). Moreover, both routing similarity and expert utilization display a pronounced layerwise structure.\u300d\u3001\u300cDominant languages serve as central hubs for cross-lingual capacity sharing, high-resource languages rely heavily on shared experts, whereas low-resource languages depend more on language- exclusive experts yet remain weak\u300d\u3068\u7d0d\u5f97\u611f\u304c\u3042\u308a\u3001\u304b\u3064\u3001\u8208\u5473\u6df1\u3044\u7d50\u679c<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/conctsai\/Multilingualism-in-Mixture-of-Experts-LLMs\">GitHub &#8211; conctsai\/Multilingualism-in-Mixture-of-Experts-LLMs<\/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":[250,267],"class_list":["post-8150","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-mixture-of-experts","tag-multilingual"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8150","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=8150"}],"version-history":[{"count":3,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8150\/revisions"}],"predecessor-version":[{"id":8164,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8150\/revisions\/8164"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8150"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8150"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8150"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}