{"id":6208,"date":"2025-02-12T06:29:00","date_gmt":"2025-02-11T21:29:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6208"},"modified":"2025-02-12T06:29:00","modified_gmt":"2025-02-11T21:29:00","slug":"taid-temporally-adaptive-interpolated-distillation-for-efficient-knowledge-transfer-in-language-models","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6208","title":{"rendered":"TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models\u00a0<\/strong>[6.8]<br>\u672c\u7a3f\u3067\u306f,\u65b0\u898f\u306a\u77e5\u8b58\u84b8\u7559\u6cd5\u3067\u3042\u308b$textitTemporally Adaptive Interpolated Distillation (TAID)$\u3092\u7d39\u4ecb\u3059\u308b\u3002 TAID\u306f,\u5404\u7a2e\u30e2\u30c7\u30eb\u30b5\u30a4\u30ba\u304a\u3088\u3073\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306b\u5bfe\u3057\u3066,\u547d\u4ee4\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3068\u4e8b\u524d\u5b66\u7fd2\u306e\u30b7\u30ca\u30ea\u30aa\u306b\u304a\u3044\u3066\u512a\u308c\u305f\u6027\u80fd\u3092\u793a\u3059\u3002 \u3053\u308c\u3089\u306e\u7d50\u679c\u306f\u3001TAID\u304c\u9ad8\u6027\u80fd\u3067\u52b9\u7387\u7684\u306a\u30e2\u30c7\u30eb\u306e\u4f5c\u6210\u306b\u6709\u52b9\u3067\u3042\u308b\u3053\u3068\u3092\u793a\u3057\u3001\u3088\u308a\u30a2\u30af\u30bb\u30b9\u3057\u3084\u3059\u3044AI\u6280\u8853\u306e\u958b\u767a\u3092\u63a8\u9032\u3057\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.16937v2\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.16937v2\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 29 Jan 2025 05:51:25 GMT)<\/li>\n\n\n\n<li>\u300cTAID reduces the gap between teacher and student model throughout the training process by dynamically introducing an intermediate teacher that interpolates teacher and student models to provide a target distribution with a modest capability\u300d\u3068\u3044\u3046\u84b8\u7559\u6cd5\u306e\u63d0\u6848<\/li>\n\n\n\n<li>\u30cb\u30e5\u30fc\u30b9\u30ea\u30ea\u30fc\u30b9\u306f<a href=\"https:\/\/sakana.ai\/taid-jp\/\">\u65b0\u624b\u6cd5\u300cTAID\u300d\u3092\u7528\u3044\u305f\u5c0f\u898f\u6a21\u65e5\u672c\u8a9e\u8a00\u8a9e\u30e2\u30c7\u30eb\u300cTinySwallow-1.5B\u300d\u306e\u516c\u958b<\/a>\u3001\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/huggingface.co\/collections\/SakanaAI\/tinyswallow-676cf5e57fff9075b5ddb7ec\">TinySwallow &#8211; a SakanaAI Collection<\/a><\/li>\n\n\n\n<li>Deepseek R1\u306e\u3088\u3046\u306b\u30e9\u30a4\u30bb\u30f3\u30b9\u4e0a\u84b8\u7559\u3092\u8a31\u53ef\u3057\u3066\u3044\u308bLRM\/LLM\u304c\u51fa\u3066\u304d\u305f\u3053\u3068\u306b\u3088\u308b\u3053\u306e\u624b\u306e\u624b\u6cd5\u306e\u91cd\u8981\u6027\u304c\u4e0a\u304c\u3063\u3066\u3044\u308b\u3088\u3046\u306b\u601d\u3046\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":[635],"class_list":["post-6208","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-635"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6208","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=6208"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6208\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6208"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6208"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6208"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}