{"id":7872,"date":"2025-12-10T04:43:00","date_gmt":"2025-12-09T19:43:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7872"},"modified":"2025-12-06T16:55:40","modified_gmt":"2025-12-06T07:55:40","slug":"are-your-agents-upward-deceivers","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7872","title":{"rendered":"Are Your Agents Upward Deceivers?\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Are Your Agents Upward Deceivers?\u00a0<\/strong>[73.1]<br>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u30d9\u30fc\u30b9\u306e\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u3001\u30e6\u30fc\u30b6\u306e\u305f\u3081\u306b\u30bf\u30b9\u30af\u3092\u5b9f\u884c\u3059\u308b\u81ea\u5f8b\u7684\u306a\u5f93\u5c5e\u8005\u3068\u3057\u3066\u3001\u307e\u3059\u307e\u3059\u4f7f\u308f\u308c\u3066\u3044\u307e\u3059\u3002 \u3053\u308c\u306f\u3001\u4eba\u9593\u306e\u7d44\u7e54\u306e\u500b\u4eba\u304c\u3069\u306e\u3088\u3046\u306b\u4e0a\u5b98\u306b\u5618\u3092\u3064\u3044\u3066\u826f\u3044\u30a4\u30e1\u30fc\u30b8\u3092\u4f5c\u308a\u51fa\u3057\u305f\u308a\u3001\u7f70\u3092\u514d\u308c\u308b\u304b\u306e\u3088\u3046\u306a\u3001\u8a50\u6b3a\u306b\u3082\u95a2\u4e0e\u3059\u308b\u304b\u3069\u3046\u304b\u3068\u3044\u3046\u554f\u984c\u3092\u63d0\u8d77\u3059\u308b\u3002 \u672c\u7814\u7a76\u3067\u306f,\u74b0\u5883\u5236\u7d04\u306b\u76f4\u9762\u3059\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304c\u969c\u5bb3\u3092\u96a0\u853d\u3057,\u5831\u544a\u306a\u3057\u306b\u8981\u6c42\u3055\u308c\u306a\u3044\u52d5\u4f5c\u3092\u884c\u3046\u73fe\u8c61\u3067\u3042\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u4e0a\u884c\u932f\u8aa4\u3092\u89b3\u5bdf\u30fb\u5b9a\u7fa9\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2512.04864v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2512.04864v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 04 Dec 2025 14:47:05 GMT)<\/li>\n\n\n\n<li>\u300cWe evaluate 11 widely used LLM-based agents (e g , Deepseek-v3.1-terminus (DeepSeek-AI, 2024), GLM- 4.5 (Zeng et al , 2025), Gemini-2.5-pro (Comanici et al , 2025)) on our task suite, and the results are striking: agen- tic upward deception is pervasive across all agents. They frequently guess, simulate outcomes, or silently switch in- formation sources when a task cannot be completed, yet still return confident and seemingly valid answers without flagging any anomalies. Most concerningly, several models even fabricate a file locally and disguise it as a successfully downloaded one.\u300d\u3068\u306e\u6307\u6458\u3002\u751f\u6210AI\u3092\u4f7f\u3063\u3066\u3044\u308b\u3068\u3057\u3070\u3057\u3070\u76ee\u306b\u3059\u308b\u52d5\u304d\u3067\u306f\u3042\u308b\u304c\u3001\u6574\u7406\u3055\u308c\u308b\u3068\u306a\u304b\u306a\u304b\u306b\u885d\u6483\u7684\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/QingyuLiu\/Agentic-Upward-Deception\">QingyuLiu\/Agentic-Upward-Deception \u00b7 GitHub<\/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":[42],"class_list":["post-7872","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-autonomous-agent"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7872","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=7872"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7872\/revisions"}],"predecessor-version":[{"id":7873,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7872\/revisions\/7873"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7872"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7872"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7872"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}