{"id":7892,"date":"2025-12-11T05:50:00","date_gmt":"2025-12-10T20:50:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7892"},"modified":"2025-12-06T17:54:14","modified_gmt":"2025-12-06T08:54:14","slug":"are-llms-good-safety-agents-or-a-propaganda-engine","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7892","title":{"rendered":"Are LLMs Good Safety Agents or a Propaganda Engine?"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Are LLMs Good Safety Agents or a Propaganda Engine?\u00a0<\/strong>[74.9]<br>PSP\u306f\u3001\u660e\u3089\u304b\u306b\u653f\u6cbb\u7684\u6587\u8108\u304b\u3089\u3001\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u62d2\u7d76\u884c\u52d5\u3092\u8abf\u3079\u308b\u305f\u3081\u306b\u69cb\u7bc9\u3055\u308c\u305f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3067\u3042\u308b\u3002 PSP\u306f\u3001\u30a4\u30f3\u30bf\u30fc\u30cd\u30c3\u30c8\u4e0a\u3067\u516c\u958b\u3055\u308c\u3066\u3044\u308b2\u3064\u306e\u30c7\u30fc\u30bf\u30bd\u30fc\u30b9\u304b\u3089\u65e2\u5b58\u306e\u691c\u95b2\u3055\u308c\u305f\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u30d5\u30a9\u30fc\u30de\u30c3\u30c8\u3059\u308b\u3053\u3068\u3067\u69cb\u7bc9\u3055\u308c\u3066\u3044\u308b\u3002 1)\u30c7\u30fc\u30bf\u99c6\u52d5\u578b(PSP\u3092\u6697\u9ed9\u5316\u3059\u308b)\u3068\u8868\u73fe\u30ec\u30d9\u30eb\u306e\u30a2\u30d7\u30ed\u30fc\u30c1(\u653f\u6cbb\u6982\u5ff5\u3092\u751f\u304b\u3057\u305f)\u306b\u3088\u308b7\u3064\u306eLSM\u306b\u304a\u3051\u308b\u653f\u6cbb\u7684\u611f\u53d7\u6027\u306e\u5f71\u97ff,2)\u30a4\u30f3\u30b8\u30a7\u30af\u30b7\u30e7\u30f3\u653b\u6483(PIA)\u306b\u3088\u308bPSP\u306b\u5bfe\u3059\u308b\u30e2\u30c7\u30eb\u8106\u5f31\u6027\u306b\u3064\u3044\u3066\u691c\u8a0e\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2511.23174v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2511.23174v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Fri, 28 Nov 2025 13:36:00 GMT)<\/li>\n\n\n\n<li>LLM\u304c\u5fdc\u7b54\u3092\u62d2\u5426\u3059\u308b\u6709\u5bb3\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u901a\u3058\u3066\u691c\u95b2\u6709\u7121\u3092\u8abf\u3079\u3088\u3046\u3068\u3044\u3046\u7814\u7a76\u3002<\/li>\n\n\n\n<li>\u300c1) some models like DeepSeek R1 and Llama 3.1 actively apply censorship policies by refusing to content that have negligible malicious intent; 2) LLMs like Llama 3.1 are quite sensitive to political contexts, in contrast to guardrail models like PromptGuard; 3) models actively enter a state of ethical dilemma\/confusion, when attacked with PIAs that contain competing objectives; 4) neither the number of parameters in a model nor the context of countries has an effect on the refusal distributions.\u300d\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":[1],"tags":[],"class_list":["post-7892","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7892","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=7892"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7892\/revisions"}],"predecessor-version":[{"id":7893,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7892\/revisions\/7893"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7892"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7892"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7892"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}