{"id":5475,"date":"2024-09-25T04:03:00","date_gmt":"2024-09-24T19:03:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5475"},"modified":"2024-09-25T04:03:00","modified_gmt":"2024-09-24T19:03:00","slug":"coca-regaining-safety-awareness-of-multimodal-large-language-models-with-constitutional-calibration","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5475","title":{"rendered":"CoCA: Regaining Safety-awareness of Multimodal Large Language Models with Constitutional Calibration\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>CoCA: Regaining Safety-awareness of Multimodal Large Language Models with Constitutional Calibration\u00a0<\/strong>[90.4]<br>\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u306a\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb (MLLM) \u306f\u3001\u8996\u899a\u7684\u306a\u5165\u529b\u3092\u542b\u3080\u4f1a\u8a71\u3078\u306e\u95a2\u4e0e\u306b\u304a\u3044\u3066\u9855\u8457\u306a\u6210\u529f\u3092\u53ce\u3081\u3066\u3044\u308b\u3002 \u8996\u899a\u7684\u30e2\u30c0\u30ea\u30c6\u30a3\u306e\u7d71\u5408\u306f\u3001MLLM\u304c\u60aa\u610f\u306e\u3042\u308b\u8996\u899a\u7684\u5165\u529b\u306b\u5f71\u97ff\u3092\u53d7\u3051\u3084\u3059\u3044\u3068\u3044\u3046\u3001\u30e6\u30cb\u30fc\u30af\u306a\u8106\u5f31\u6027\u3092\u5c0e\u5165\u3057\u3066\u3044\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u51fa\u529b\u5206\u5e03\u3092\u6821\u6b63\u3059\u308b\u3053\u3068\u3067MLLM\u306e\u5b89\u5168\u6027\u3092\u5411\u4e0a\u3059\u308bCoCA\u6280\u8853\u3092\u7d39\u4ecb\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2409.11365v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2409.11365v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 17 Sep 2024 17:14:41 GMT)<\/li>\n\n\n\n<li>MLLM\u306b\u304a\u3044\u3066\u60aa\u610f\u306e\u3042\u308b\u753b\u50cf\u3092\u4ecb\u3057\u305f\u653b\u6483\u304c\u554f\u984c\u306b\u306a\u308b\u304c\u3001\u305d\u306e\u5bfe\u5fdc\u306b\u95a2\u3059\u308b\u8ad6\u6587\u3002<\/li>\n\n\n\n<li>\u300cWe first make the observation that despite the integration of visual modality makes the MLLMs more vulnerable, the inherent safetyawareness of MLLMs still exists.\u300d\u306f\u3078\u30fc\u3068\u3044\u3046\u611f\u3058\u3001<\/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":[32,251],"class_list":["post-5475","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-attack","tag-mllm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5475","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=5475"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5475\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5475"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5475"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5475"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}