{"id":8450,"date":"2026-05-22T02:41:28","date_gmt":"2026-05-21T18:41:28","guid":{"rendered":"https:\/\/infernews.com\/blog\/videorlvr\/"},"modified":"2026-05-22T02:41:28","modified_gmt":"2026-05-21T18:41:28","slug":"videorlvr","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/videorlvr\/","title":{"rendered":"VideoRLVR\uff1a\u6559\u5f71\u7247\u6a21\u578b\u5b78\u6703\u63a8\u7406"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/05\/pasted-208d2ad8e415.jpg\" alt=\"Repository image for luka-group\/VideoRLVR\"><\/figure>\n<p>VideoRLVR \u662f\u4e00\u500b\u7528\u4f86\u8a13\u7df4\u5f71\u7247\u63a8\u7406\u6a21\u578b\u7684\u9805\u76ee\uff0c\u6838\u5fc3\u505a\u6cd5\u662f\u628a\u5f37\u5316\u5b78\u7fd2\u52a0\u5165\u5f71\u7247\u751f\u6210\u6d41\u7a0b\uff0c\u4e26\u7528\u300c\u53ef\u9a57\u8b49\u300d\u7684\u734e\u52f5\u4f86\u5224\u65b7\u7b54\u6848\u662f\u5426\u6b63\u78ba\u3002\u7c21\u55ae\u8aaa\uff0c\u5b83\u4e0d\u662f\u53ea\u8ffd\u6c42\u756b\u9762\u50cf\u771f\uff0c\u800c\u662f\u5e0c\u671b\u6a21\u578b\u5728\u751f\u6210\u5f71\u7247\u6642\uff0c\u9023\u5e36\u8868\u73fe\u51fa\u53ef\u6aa2\u67e5\u7684\u89e3\u984c\u80fd\u529b\u3002<\/p>\n<p>\u9019\u500b\u9805\u76ee\u76ee\u524d\u570d\u7e5e Wan2.2-TI2V-5B \u5c55\u958b\uff0c\u4e26\u4ee5 Maze\u3001FlowFree\u3001Sokoban \u9019\u985e\u6709\u660e\u78ba\u898f\u5247\u7684\u4efb\u52d9\u4f5c\u70ba\u8a13\u7df4\u8207\u8a55\u4f30\u5834\u666f\u3002\u9019\u985e\u8a2d\u8a08\u7684\u597d\u8655\uff0c\u662f\u6a21\u578b\u8868\u73fe\u4e0d\u53ea\u9760\u4e3b\u89c0\u89c0\u611f\uff0c\u800c\u662f\u53ef\u4ee5\u900f\u904e\u4efb\u52d9\u6210\u529f\u8207\u5426\u4f86\u91cf\u5ea6\uff0c\u5c0d\u7814\u7a76\u63a8\u7406\u80fd\u529b\u7279\u5225\u91cd\u8981\u3002<\/p>\n<p>\u8981\u7406\u89e3\u9019\u500b\u9805\u76ee\uff0c\u53ef\u5148\u7531\u5b83\u63d0\u4f9b\u7684\u8cc7\u6e90\u5165\u624b\uff1a\u516c\u958b\u96c6\u5408\u5167\u6709 SFT \u8207 RLVR \u6aa2\u67e5\u9ede\uff0c\u4e5f\u6709\u8a13\u7df4\u53ca\u6e2c\u8a66\u8cc7\u6599\u3002\u7a0b\u5f0f\u7d50\u69cb\u4ea6\u5206\u958b\u4e86\u8a13\u7df4\u3001\u63a8\u8ad6\u8207\u8a55\u4f30\u8173\u672c\uff0c\u4e26\u63d0\u4f9b\u591a\u4efb\u52d9\u53ca\u55ae\u4e00\u4efb\u52d9\u7248\u672c\uff0c\u65b9\u4fbf\u6bd4\u8f03\u4e0d\u540c\u8a2d\u5b9a\u4e0b\u7684\u7d50\u679c\u3002<\/p>\n<ul>\n<li>\u91cd\u9ede\u653e\u5728<strong>\u53ef\u9a57\u8b49\u734e\u52f5<\/strong>\uff0c\u6bd4\u53ea\u770b\u4e3b\u89c0\u751f\u6210\u8cea\u7d20\u66f4\u6613\u8a55\u4f30<\/li>\n<li>\u4ee5 Wan2.2-TI2V-5B \u70ba\u57fa\u790e\uff0c\u63d0\u4f9b SFT \u8207 RLVR \u76f8\u95dc\u6a21\u578b<\/li>\n<li>\u4efb\u52d9\u6db5\u84cb Maze\u3001FlowFree\u3001Sokoban\uff0c\u504f\u5411\u898f\u5247\u6e05\u6670\u7684\u63a8\u7406\u6e2c\u8a66<\/li>\n<li>\u5df2\u6574\u7406\u6a21\u578b\u8207\u8cc7\u6599\u96c6\u5230 Hugging Face\uff0c\u67e5\u627e\u8cc7\u6e90\u8f03\u65b9\u4fbf<\/li>\n<\/ul>\n<p>\u9019\u500b\u9805\u76ee\u7684\u65b0\u610f\uff0c\u5728\u65bc\u628a\u5f71\u7247\u751f\u6210\u8207\u53ef\u91cf\u5316\u7684\u63a8\u7406\u8a13\u7df4\u66f4\u7dca\u5bc6\u5730\u7d50\u5408\uff0c\u4e26\u4e14\u660e\u78ba\u63d0\u4f9b\u4e00\u5957\u53ef\u91cd\u8907\u7684\u8a13\u7df4\u914d\u65b9\u3002\u5f9e README \u53ef\u898b\uff0c\u5b83\u4ea6\u5305\u542b\u591a\u4efb\u52d9\u8a13\u7df4\u3001\u7d14\u6210\u529f\u8a0a\u865f\u7248\u672c\uff0c\u4ee5\u53ca OOD \u63a8\u8ad6\u8207\u8a55\u4f30\u8173\u672c\uff0c\u8868\u793a\u4f5c\u8005\u4e0d\u53ea\u95dc\u5fc3\u662f\u5426\u5b78\u6703\u6307\u5b9a\u984c\u76ee\uff0c\u4e5f\u5728\u610f\u6a21\u578b\u96e2\u958b\u719f\u6089\u5206\u4f48\u5f8c\u7684\u8868\u73fe\u3002<\/p>\n<p>\u9069\u5408\u7559\u610f\u9019\u500b\u9805\u76ee\u7684\u4eba\uff0c\u5305\u62ec\u7814\u7a76\u591a\u6a21\u614b\u6a21\u578b\u3001\u5f71\u7247\u751f\u6210\u3001\u5f37\u5316\u5b78\u7fd2\uff0c\u6216\u60f3\u89c0\u5bdf\u6a21\u578b\u5982\u4f55\u5728\u898f\u5247\u74b0\u5883\u4e2d\u8868\u73fe\u63a8\u7406\u80fd\u529b\u7684\u958b\u767c\u8005\u3002\u81f3\u65bc\u6027\u80fd\u6578\u5b57\uff0c\u9019\u4efd\u5132\u5b58\u5eab\u6458\u8981\u672a\u5c55\u793a\u5b8c\u6574\u91cf\u5316\u7d50\u679c\uff0c\u56e0\u6b64\u8f03\u7a69\u59a5\u7684\u770b\u6cd5\u662f\uff1a\u5b83\u7684\u50f9\u503c\u76ee\u524d\u66f4\u504f\u5411\u7814\u7a76\u65b9\u6cd5\u3001\u8a13\u7df4\u6d41\u7a0b\u8207\u516c\u958b\u8cc7\u6e90\uff0c\u800c\u4e0d\u662f\u55ae\u9760\u4e00\u5169\u500b\u5206\u6578\u5b9a\u52dd\u8ca0\u3002<\/p>\n<p><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/luka-group\/VideoRLVR\" rel=\"noopener noreferrer\">https:\/\/github.com\/luka-group\/VideoRLVR<\/a><\/p>\n<p><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2605.15458\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2605.15458<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9019\u500b\u9805\u76ee\u628a\u5f37\u5316\u5b78\u7fd2\u5e36\u5165\u5f71\u7247\u63a8\u7406\u8a13\u7df4\uff0c\u91cd\u9ede\u662f\u7528\u53ef\u9a57\u8b49\u734e\u52f5\u63d0\u5347\u89e3\u984c\u80fd\u529b\u3002\u5c0d\u60f3\u7814\u7a76\u751f\u6210\u5f0f\u5f71\u7247\u8207\u63a8\u7406\u7d50\u5408\u7684\u4eba\u7279\u5225\u503c\u5f97\u7559\u610f\u3002<\/p>\n","protected":false},"author":8,"featured_media":8449,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[133,173,160,149],"tags":[],"class_list":["post-8450","post","type-post","status-publish","format-standard","hentry","category-133","category-173","category-160","category-149"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8450","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=8450"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8450\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8449"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8450"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8450"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8450"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}