zhengkid/Dream-RSI

&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; <img width="721" src="assets/tagline-light.svg" [](papers/Dream-RSI.pdf) [](https://www.dream-rsi.com) [](https://www.dream-rsi.com) [](papers/Dream-RSI.pdf) Tong Zheng1,2, Xidong Wu1✉, Zheng Zhang1✉, Zhankui He3, Chaoyi Zhang1, Benjamin Coleman3, Ruoqiao Wei1, Di Bai3, Haolin Liu4, Rui Liu2, Xue Wang1, Yue Zhuan1, Wang-Cheng Kang3, Renkai Xiang1, Heng Huang2, Xinwu Cheng1, Yunsong Guo1 1Google &nbsp;·&nbsp; 2University of Maryland, College Park &nbsp;·&nbsp; 3Google DeepMind &nbsp;·&nbsp; 4University of Virginia ✉ Corresponding authors google-research/dream-rsi. paper (PDF) · dream-rsi.com Progress in recursive self-improvement hinges on effective exploration. As self-improvement targets get harder, discovery stretches over thousands of proposal–evaluation cycles, and poor exploration wastes substantial computation on ineffective search directions. Managing and improving the exploration strategy is the bottleneck, and current systems face a dilemma: fixed strategies cannot adapt as search spaces scale, while optimizing the policy online means navigating a vast meta-search space under delayed and expensive feedback — assessing one exploration policy requires observing how it shapes an entire discovery process. Our…