{"id":5968,"date":"2025-01-03T03:28:00","date_gmt":"2025-01-02T18:28:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5968"},"modified":"2025-01-03T03:28:00","modified_gmt":"2025-01-02T18:28:00","slug":"deliberation-in-latent-space-via-differentiable-cache-augmentation","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5968","title":{"rendered":"Deliberation in Latent Space via Differentiable Cache Augmentation\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Deliberation in Latent Space via Differentiable Cache Augmentation\u00a0<\/strong>[48.2]<br>\u51cd\u7d50\u3057\u305f\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u3092\u30aa\u30d5\u30e9\u30a4\u30f3\u30b3\u30d7\u30ed\u30bb\u30c3\u30b5\u3067\u62e1\u5f35\u3057,\u30ad\u30fc\u5024(kv)\u30ad\u30e3\u30c3\u30b7\u30e5\u3067\u52d5\u4f5c\u3059\u308b\u3053\u3068\u3092\u793a\u3059\u3002 \u3053\u306e\u30b3\u30d7\u30ed\u30bb\u30c3\u30b5\u306f\u3001\u5f8c\u7d9a\u306e\u5fa9\u53f7\u5316\u306e\u5fe0\u5b9f\u6027\u3092\u6539\u5584\u3059\u308b\u305f\u3081\u306b\u8a2d\u8a08\u3055\u308c\u305f\u9045\u5ef6\u57cb\u3081\u8fbc\u307f\u306e\u30bb\u30c3\u30c8\u3067\u30ad\u30e3\u30c3\u30b7\u30e5\u3092\u5897\u5f37\u3059\u308b\u3002 \u30ad\u30e3\u30c3\u30b7\u30e5\u304c\u62e1\u5f35\u3055\u308c\u308b\u3068\u3001\u30c7\u30b3\u30fc\u30c0\u306f\u591a\u6570\u306e\u30c8\u30fc\u30af\u30f3\u306b\u5bfe\u3057\u3066\u4f4e\u3044\u30d1\u30fc\u30d7\u30ec\u30ad\u30b7\u30c6\u30a3\u3092\u9054\u6210\u3067\u304d\u308b\u3053\u3068\u3092\u793a\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2412.17747v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2412.17747v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 23 Dec 2024 18:02:25 GMT)<\/li>\n\n\n\n<li>\u300cThis paper introduces differentiable cache augmentation, a novel method for enhancing frozen decoderonly language models by incorporating a learned coprocessor that operates on the model\u2019s kv-cache.\u300d\u3068\u3044\u3046\u63d0\u6848\u3002coprocessor\u306f\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u53ef\u80fd\u3002<\/li>\n\n\n\n<li>\u8ad6\u6587\u3067\u3082\u8a00\u53ca\u306e\u3042\u308b\u00a0Pause Token\u00a0\u3068\u8003\u3048\u65b9\u306f\u4f3c\u3066\u3044\u308b\u304c\u3001\u3053\u306e\u624b\u6cd5\u306e\u307b\u3046\u304c\u3088\u308a\u5f37\u529b\u3068\u306e\u3053\u3068\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":[223],"class_list":["post-5968","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5968","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=5968"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5968\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5968"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5968"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5968"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}