{"id":6564,"date":"2025-04-15T06:05:00","date_gmt":"2025-04-14T21:05:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6564"},"modified":"2025-04-15T06:05:00","modified_gmt":"2025-04-14T21:05:00","slug":"sota-with-less-mcts-guided-sample-selection-for-data-efficient-visual-reasoning-self-improvement","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6564","title":{"rendered":"SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement\u00a0<\/strong>[100.9]<br>ThinkLite-VL\u306fQwen2.5-VL-7B\u30a4\u30f3\u30b9\u30c8\u30e9\u30af\u30b7\u30e7\u30f3\u306e\u5e73\u5747\u6027\u80fd\u30927%\u5411\u4e0a\u3055\u305b\u308b\u3002 \u79c1\u305f\u3061\u306e\u30b3\u30fc\u30c9\u3001\u30c7\u30fc\u30bf\u3001\u30e2\u30c7\u30eb\u306fhttps:\/\/github.com\/si0wang\/ThinkLite-VL.org\u3067\u516c\u958b\u3055\u308c\u3066\u3044\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2504.07934v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2504.07934v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 10 Apr 2025 17:49:05 GMT)<\/li>\n\n\n\n<li>\u52b9\u7387\u306e\u3088\u3044Vision-Language\u30e2\u30c7\u30eb\u306e\u63a8\u8ad6\u5f37\u5316\u65b9\u6cd5\u306e\u63d0\u6848\u3002\u300cOur model achieves SoTA performance using only 11k data, and without any additional knowledge distillation.\u300d\u3068\u4f7f\u7528\u30c7\u30fc\u30bf\u304c\u5c11\u306a\u3044\u3002\u30ab\u30ae\u306f\u30c7\u30fc\u30bf\u54c1\u8cea\u3068\u306e\u3053\u3068\u300cOur key insight highlights the critical importance of selecting genuinely challenging examples for Reinforcement Fine-Tuning (RFT).\u300d<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/si0wang\/ThinkLite-VL\">GitHub &#8211; si0wang\/ThinkLite-VL<\/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":[102,232,251],"class_list":["post-6564","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-dataset-distillation","tag-lrm","tag-mllm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6564","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=6564"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6564\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6564"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6564"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6564"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}