The Path Not Taken: RLVR Provably Learns Off the Principals
机构 * Meta AI ; The University of Texas at Austin(德克萨斯大学奥斯汀分校)
Comments Preliminary version accepted as a spotlight in NeurIPS 2025 Workshop on Efficient Reasoning
高校专区
机构 * Meta AI ; The University of Texas at Austin(德克萨斯大学奥斯汀分校)
Comments Preliminary version accepted as a spotlight in NeurIPS 2025 Workshop on Efficient Reasoning
机构 * Sony Interactive Entertainment(索尼互动娱乐) ; University of Texas at Austin(德克萨斯大学奥斯汀分校)
机构 * The University of Texas at Austin(德克萨斯大学奥斯汀分校)
机构 * The State University of New York at Binghamton(纽约州立大学布林顿分校) ; The University of Texas at Austin(德克萨斯大学奥斯汀分校) ; Sony AI(索尼人工智能) ; Hello Robot(Hello Robot公司) ; Shanghai AI Laboratory(上海人工智能实验室) ; OneStar Robotics(OneStar机器人)
Journal ref The International Journal of Robotics Research, 2025, Vol. 0(0), pp. 1-19