Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learning

作者：Patrick Yin, Tyler Westenbroek, Octi Zhang, Joshua Tran, Ignacio Dagnino, Eeshani Shilamkar, Numfor Mbiziwo-Tiapo, Simran Bagaria, Xinlei Liu, Galen Mullins, Andrey Kolobov, Abhishek Gupta · 单位：University of Washington, NVIDIA, Microsoft Research · 会议/期刊：ICLR 2026 · 日期：2026-03-16 · 来源：Low-Level Learning-Based Action Modelling / Learning Strategy / Model-free Reinforcement Learning / RL in Pre-Training