{"id":4673,"date":"2024-04-10T05:12:00","date_gmt":"2024-04-09T20:12:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4673"},"modified":"2024-04-10T05:12:00","modified_gmt":"2024-04-09T20:12:00","slug":"on-the-multilingual-ability-of-decoder-based-pre-trained-language-models-finding-and-controlling-language-specific-neurons","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4673","title":{"rendered":"On the Multilingual Ability of Decoder-based Pre-trained Language Models: Finding and Controlling Language-Specific Neurons"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>On the Multilingual Ability of Decoder-based Pre-trained Language Models: Finding and Controlling Language-Specific Neurons\u00a0<\/strong>[37.3]<br>\u591a\u8a00\u8a9e\u30c7\u30b3\u30fc\u30c0\u3092\u7528\u3044\u305f\u8a00\u8a9e\u30e2\u30c7\u30eb(PLM)\u306e\u30cb\u30e5\u30fc\u30ed\u30f3\u30ec\u30d9\u30eb\u306e\u5185\u90e8\u6319\u52d5\u306e\u89e3\u6790 \u8a00\u8a9e\u56fa\u6709\u306e\u30cb\u30e5\u30fc\u30ed\u30f3\u306f\u3001\u8a00\u8a9e\u9593\u3067\u308f\u305a\u304b\u306b\u91cd\u306a\u308a(5%)\u3001\u30e6\u30cb\u30fc\u30af\u3067\u3042\u308b\u3053\u3068\u3092\u793a\u3059\u3002 \u63a8\u8ad6\u4e2d\u306b\u5404\u30e2\u30c7\u30eb\u306b\u304a\u3051\u308b\u5168\u30cb\u30e5\u30fc\u30ed\u30f3\u306e1%\u672a\u6e80\u3092\u30bf\u30f3\u30d1\u3057\u3001\u5c11\u6570\u306e\u8a00\u8a9e\u7279\u7570\u7684\u30cb\u30e5\u30fc\u30ed\u30f3\u3068\u306e\u30bf\u30f3\u30d1\u30ea\u30f3\u30b0\u304c\u30c6\u30ad\u30b9\u30c8\u751f\u6210\u306b\u304a\u3051\u308b\u30bf\u30fc\u30b2\u30c3\u30c8\u8a00\u8a9e\u767a\u751f\u306e\u78ba\u7387\u3092\u5287\u7684\u306b\u5909\u5316\u3055\u305b\u308b\u3053\u3068\u3092\u5b9f\u8a3c\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2404.02431v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2404.02431v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 03 Apr 2024 03:37:22 GMT)<\/li>\n\n\n\n<li>PLM\u306b\u304a\u3051\u308b\u591a\u8a00\u8a9e\u6027\u306e\u5206\u6790\u3001\u300cThe experimental results demonstrate that language-specific neurons mainly exist in the first and last few layers, regardless of the language, model size, and model variants.\u300d\u3068\u3044\u3046Findings\u306f<a href=\"https:\/\/devneko.jp\/wordpress\/?p=4505\">Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Models \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb (devneko.jp)<\/a>\u306a\u3069\u4ed6\u306e\u7d50\u679c\u3068\u6574\u5408\u7684\u3067\u3042\u308b\u3088\u3046\u306b\u601d\u3048\u308b\u3002Controlling Language-specific Neurons\u3067\u306e\u300cIn other words, the desired language could be generated by intentionally igniting target neurons.\u300d\u306f\u9762\u767d\u3044\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":[267],"class_list":["post-4673","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-multilingual"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4673","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=4673"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4673\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4673"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4673"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4673"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}