{"id":6235,"date":"2025-02-21T06:17:00","date_gmt":"2025-02-20T21:17:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6235"},"modified":"2025-02-21T06:17:00","modified_gmt":"2025-02-20T21:17:00","slug":"selfcite-self-supervised-alignment-for-context-attribution-in-large-language-models","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6235","title":{"rendered":"SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models\u00a0<\/strong>[51.9]<br>SelfCite\u306f\u81ea\u5df1\u6559\u5e2b\u578b\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u3067\u3001LLM\u3092\u8abf\u6574\u3057\u3066\u3001\u751f\u6210\u3055\u308c\u305f\u5fdc\u7b54\u306b\u304a\u3051\u308b\u6587\u306b\u5bfe\u3059\u308b\u9ad8\u54c1\u8cea\u3067\u304d\u3081\u7d30\u304b\u3044\u6587\u30ec\u30d9\u30eb\u306e\u5f15\u7528\u3092\u751f\u6210\u3059\u308b\u3002 \u30b3\u30b9\u30c8\u3068\u52b4\u50cd\u96c6\u7d04\u7684\u306a\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u306b\u983c\u308b\u4ee3\u308f\u308a\u306b\u3001SelfCite\u306fLLM\u81ea\u4f53\u304c\u63d0\u4f9b\u3059\u308b\u5831\u916c\u30b7\u30b0\u30ca\u30eb\u3092\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30a2\u30d6\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u901a\u3058\u3066\u6d3b\u7528\u3059\u308b\u3002 SelfCite\u306e\u6709\u52b9\u6027\u306f\u30015\u3064\u306e\u9577\u6587\u8cea\u554f\u5fdc\u7b54\u30bf\u30b9\u30af\u306b\u308f\u305f\u308bLongBench-Cite\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306b\u304a\u3044\u3066\u3001\u5f15\u7528F1\u30925.3\u30dd\u30a4\u30f3\u30c8\u307e\u3067\u5897\u3084\u3059\u3053\u3068\u306b\u3088\u3063\u3066\u793a\u3055\u308c\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2502.09604v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2502.09604v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 13 Feb 2025 18:55:13 GMT)<\/li>\n\n\n\n<li>\u300cFirst, the full context is used to generate a response. Then, the framework evaluates the probability of generating the same response after (1) removing the cited sentences from the context and (2) using only the cited sentences in the context. The probability drop and hold are computed from these probability differences, and their sum is used as the final reward.\u300d\u3068\u3044\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u306ereward\u8a08\u7b97\u3001preference optimization with SimPO\u304c\u826f\u3044\u7d50\u679c\u3060\u3063\u305f\u3068\u306e\u5831\u544a\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":[64],"class_list":["post-6235","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-citation"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6235","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=6235"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6235\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6235"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6235"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6235"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}