{"id":5140,"date":"2024-07-16T05:14:00","date_gmt":"2024-07-15T20:14:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5140"},"modified":"2024-07-16T05:14:00","modified_gmt":"2024-07-15T20:14:00","slug":"copybench-measuring-literal-and-non-literal-reproduction-of-copyright-protected-text-in-language-model-generation","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5140","title":{"rendered":"CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation\u00a0<\/strong>[132.0]<br>\u8a00\u8a9e\u30e2\u30c7\u30eb(LM)\u306b\u3088\u308b\u8457\u4f5c\u6a29\u4fdd\u8b77\u3055\u308c\u305f\u30b3\u30f3\u30c6\u30f3\u30c4\u306e\u518d\u751f\u306e\u5ea6\u5408\u3044\u3092\u8a55\u4fa1\u3059\u308b\u3053\u3068\u306f\u3001AI\u3068\u6cd5\u7684\u306a\u30b3\u30df\u30e5\u30cb\u30c6\u30a3\u306b\u3068\u3063\u3066\u91cd\u8981\u306a\u95a2\u5fc3\u4e8b\u3067\u3042\u308b\u3002 LM\u4e16\u4ee3\u306b\u304a\u3051\u308b\u30ea\u30c6\u30e9\u30eb\u30b3\u30d4\u30fc\u3068\u975e\u30ea\u30c6\u30e9\u30eb\u30b3\u30d4\u30fc\u306e\u4e21\u65b9\u3092\u6e2c\u5b9a\u3059\u308b\u305f\u3081\u306b\u8a2d\u8a08\u3055\u308c\u305f\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308bCopyBench\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u30ea\u30c6\u30e9\u30eb\u8907\u5199\u306f\u6bd4\u8f03\u7684\u7a00\u3067\u3042\u308b\u304c\u3001\u30a4\u30d9\u30f3\u30c8\u8907\u5199\u3068\u6587\u5b57\u8907\u5199\u3068\u3044\u30462\u7a2e\u985e\u306e\u975e\u30ea\u30c6\u30e9\u30eb\u8907\u5199\u306f\u30017B\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u30e2\u30c7\u30eb\u3067\u3082\u767a\u751f\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2407.07087v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2407.07087v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 09 Jul 2024 17:58:18 GMT)<\/li>\n\n\n\n<li>\uff08\u30b7\u30f3\u30d7\u30eb\u306a\uff09Literal copying\u3060\u3051\u3067\u306a\u304f\u30d1\u30e9\u30d5\u30a7\u30fc\u30b8\u30f3\u30b0\u306a\u3069\u3092\u901a\u3057\u3066\u884c\u308f\u308c\u308bNon-literal copying\u3082\u5bfe\u8c61\u306b\u3057\u305f\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u63d0\u6848\u3068\u691c\u8a3c\u3002\u30ea\u30c6\u30e9\u30eb\u30b3\u30d4\u30fc\u304c\u5927\u898f\u6a21\u30e2\u30c7\u30eb\u3067\u554f\u984c\u306b\u306a\u308b\u306e\u306f\u76f4\u89b3\u7684\u306b\u660e\u3089\u304b\u3068\u3057\u3066\u300cIn proprietary models, the transition from GPT-3.5 to GPT-4 interestingly reduces literal copying but increases non-literal copying.\u300d\u3084\u300cwe find that current inference-time mitigation methods, although effective at reducing literal copying, are insufficient for addressing nonliteral copying.\u300d\u3068\u3044\u3046\u8a18\u8f09\u306f\u8208\u5473\u6df1\u3044\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/chentong0\/copy-bench\">GitHub &#8211; chentong0\/copy-bench: CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation<\/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":[622],"class_list":["post-5140","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-622"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5140","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=5140"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5140\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5140"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5140"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5140"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}