{"id":8370,"date":"2026-03-30T04:20:00","date_gmt":"2026-03-29T19:20:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8370"},"modified":"2026-03-29T21:26:22","modified_gmt":"2026-03-29T12:26:22","slug":"claudini-autoresearch-discovers-state-of-the-art-adversarial-attack-algorithms-for-llms","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8370","title":{"rendered":"Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs\u00a0<\/strong>[75.6]<br>\u6211\u3005\u306f,Claude Code\u3092\u30d9\u30fc\u30b9\u3068\u3057\u305femphautoresearch\u30b9\u30bf\u30a4\u30eb\u306e\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3 citepkarpathy2026autoresearch \u304c,\u65b0\u3057\u3044\u30db\u30ef\u30a4\u30c8\u30dc\u30c3\u30af\u30b9\u5bfe\u9006\u653b\u6483\u3092\u767a\u898b\u3067\u304d\u308b\u3053\u3068\u3092\u793a\u3057\u305f\u3002 textbfsignant\u306f\u3001jailbreak\u3067\u65e2\u5b58\u306e\u3059\u3079\u3066\u306e(30\u4ee5\u4e0a\u306e\u30e1\u30bd\u30c3\u30c9)\u3092\u8457\u3057\u304f\u4e0a\u56de\u308a\u3001\u30a4\u30f3\u30b8\u30a7\u30af\u30b7\u30e7\u30f3\u8a55\u4fa1\u3092\u4fc3\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2603.24511v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2603.24511v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 25 Mar 2026 16:50:56 GMT)<\/li>\n\n\n\n<li>\u300cWe show that an\u00a0<em><a href=\"https:\/\/github.com\/karpathy\/autoresearch\">autoresearch<\/a><\/em>-style pipeline powered by Claude Code discovers novel white-box adversarial attack\u00a0<em>algorithms<\/em>\u00a0that\u00a0<strong>significantly outperform<\/strong>\u00a0all existing\u00a0<a href=\"https:\/\/github.com\/romovpa\/claudini\/blob\/main\/claudini\/methods\/original\/README.md\">methods<\/a>\u00a0in jailbreaking and prompt injection evaluations.\u300d\u3068\u306e\u3053\u3068\u3002\u3002\u3002<a href=\"https:\/\/github.com\/romovpa\/claudini\/blob\/main\/.claude\/skills\/claudini\/SKILL.md\">claudini\/.claude\/skills\/claudini\/SKILL.md at main \u00b7 romovpa\/claudini \u00b7 GitHub<\/a>\u306e\u3088\u3046\u306aSKILL\u30d9\u30fc\u30b9\u3068\u3044\u3046\u306e\u3082\u4eca\u98a8\uff08\u305d\u3057\u3066\u6016\u3044\uff09\u3002<\/li>\n\n\n\n<li><a href=\"https:\/\/fugumt.com\/autoreport\/daily\/2026-03-25-2603.24511v1_Claudini_Autoresearch_Discovers_State-of-the-Art_Adversarial_Attack_Algorithms_f.html\">Daily Report 2026-03-25\u81ea\u52d5\u4f5c\u6210\u30ec\u30dd\u30fc\u30c8\uff1aClaudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs<\/a><\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/romovpa\/claudini\">GitHub &#8211; romovpa\/claudini: Autoresearch for LLM adversarial attacks \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":[352],"class_list":["post-8370","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-security"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8370","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=8370"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8370\/revisions"}],"predecessor-version":[{"id":8371,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8370\/revisions\/8371"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8370"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8370"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8370"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}