Authors: Tong Zheng, Xidong Wu, Zheng Zhang, Zhankui He, Chaoyi Zhang, Benjamin Coleman, Ruoqiao Wei, Di Bai, Haolin Liu, Rui Liu, Xue Wang, Yue Zhuan, Wang-Cheng Kang, Renkai Xiang, Heng Huang, Xinwu Cheng, Yunsong Guo

Dream-RSI introduces a framework for scalable recursive self-improvement in autonomous AI agents. The system addresses the exploration bottleneck by making discovery history explicit and programmable through a lightweight orchestration layer. The key insight utilizes accumulated discovery history as a replay simulator over the realized search space. By performing dreaming in this historical replay simulator, Dream-RSI secures immediate, low-cost off-policy feedback to evaluate and refine exploration policies without expensive online evaluations. The improved policy is redeployed online to drive further discovery, continuously expanding the simulator pool in a self-improving loop. Across algorithm engineering, mathematical optimization, and GPU kernel engineering, Dream-RSI achieves competitive or improved discovery quality while substantially reducing discovery cost in several settings. The submission originates from Tong Zheng and is available as a PDF and experimental HTML version.