{"id":7617,"date":"2025-10-20T02:55:00","date_gmt":"2025-10-19T17:55:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7617"},"modified":"2025-10-18T14:57:42","modified_gmt":"2025-10-18T05:57:42","slug":"qwen3guard-technical-report","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7617","title":{"rendered":"Qwen3Guard Technical Report"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Qwen3Guard Technical Report\u00a0<\/strong>[127.7]<br>Qwen3Guard\u306f\u3001\u591a\u8a00\u8a9e\u5b89\u5168\u30ac\u30fc\u30c9\u30ec\u30fc\u30eb\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u751f\u6210\u7684Qwen3Guard\u306f\u3001\u304d\u3081\u7d30\u304b\u3044\u4e09\u7d1a\u5224\u5b9a\u3092\u53ef\u80fd\u306b\u3059\u308b\u547d\u4ee4\u8ffd\u5f93\u30bf\u30b9\u30af\u3068\u3057\u3066\u5b89\u5168\u5206\u985e\u3092\u30ad\u30e3\u30b9\u30c8\u3059\u308b\u3002 Stream Qwen3Guard\u306f\u3001\u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u5b89\u5168\u76e3\u8996\u306e\u305f\u3081\u306e\u30c8\u30fc\u30af\u30f3\u30ec\u30d9\u30eb\u306e\u5206\u985e\u30d8\u30c3\u30c9\u3092\u5c0e\u5165\u3057\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2510.14276v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2510.14276v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 16 Oct 2025 04:00:18 GMT)<\/li>\n\n\n\n<li>\u300cwe present Qwen3Guard, a series of multilingual safety guardrail models with two specialized variants: Generative Qwen3Guard, which casts safety classification as an instruction-following task to enable fine-grained tri-class judgments (safe, controversial, unsafe); and Stream Qwen3Guard, which introduces a token-level classification head for real-time safety monitoring during incremental text generation. Both variants are available in three sizes (0.6B, 4B, and 8B parameters) and support up to 119 languages and dialects, providing comprehensive, scalable, and low-latency safety moderation for global LLM deployments.\u300d\u3068Qwen3\u30d9\u30fc\u30b9\u306e\u30ac\u30fc\u30c9\u30ec\u30fc\u30eb\u3082\u30e2\u30c7\u30eb\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/QwenLM\/Qwen3Guard\">GitHub &#8211; QwenLM\/Qwen3Guard: Qwen3Guard is a multilingual guardrail model series developed by the Qwen team at Alibaba Cloud.<\/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":[704],"class_list":["post-7617","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-704"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7617","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=7617"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7617\/revisions"}],"predecessor-version":[{"id":7618,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7617\/revisions\/7618"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7617"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7617"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7617"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}