{"id":7551,"date":"2025-10-06T05:45:00","date_gmt":"2025-10-05T20:45:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7551"},"modified":"2025-10-04T21:46:47","modified_gmt":"2025-10-04T12:46:47","slug":"ferret-ui-lite-lessons-from-building-small-on-device-gui-agents","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7551","title":{"rendered":"Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents\u00a0<\/strong>[79.8]<br>Ferret-UI Lite\u306f\u3001\u69d8\u3005\u306a\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u3067\u52d5\u4f5c\u3059\u308b\u3001\u30b3\u30f3\u30d1\u30af\u30c8\u3067\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306eGUI\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3067\u3042\u308b\u3002 Ferret-UI Lite\u306f\u3001\u4ed6\u306e\u5c0f\u898f\u6a21GUI\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3068\u306e\u7af6\u5408\u6027\u80fd\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.26539v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.26539v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 30 Sep 2025 17:13:56 GMT)<\/li>\n\n\n\n<li>Apple\u306b\u3088\u308bGUI\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u5831\u544a\u3001\u300cIn this work, we present Ferret-UI Lite, a 3B multimodal LLM designed for GUI agentic tasks with a focus on lightweight, on-device settings. Through real and synthetic data curation, inference-time visual tool use, and a two-stage SFT\u2013RL training strategy, Ferret-UI Lite achieves competitive grounding and navigation performance relative to larger models.\u300d\u3068\u5c0f\u578b\u306e\u30e2\u30c7\u30eb\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":[2],"tags":[181],"class_list":["post-7551","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-gui-agent"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7551","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=7551"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7551\/revisions"}],"predecessor-version":[{"id":7552,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7551\/revisions\/7552"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7551"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7551"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7551"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}