arXiv：cs.AI· Leijun Zhou, Yanke Yu, Zhihao Liu, Qu Xiang, Tiancheng Hu, Yiru Dong, Yihan Wu, Yangxi Xiang, Xuejun Wu, Xi Chen, Buyue Qian, Yaowei Zheng, Junhao Hu·· 16 小时前
从现有方法到新研究思路：递归自我改进的两个视角
From Existing Methods to New Research Ideas: Two Perspectives on Recursive Self-Improvement
AI 导读
一篇 arXiv 综述从两个视角梳理递归自我改进（RSI）研究：一是提出覆盖更新目标、更新组织、更新调度与范围、反馈与选择四大组共九个设计维度的统一分类体系，将现有方法置于共享设计空间；二是分析现有 RSI 方法如何借鉴优化、正则化、数据增强、持续学习和对抗学习等传统深度学习机制，并从进化计算中拓展新思路，同时回顾在线、离线及离线转在线三种设定下的自改进评测基准。
正文
Authors:Leijun Zhou, Yanke Yu, Zhihao Liu, Qu Xiang, Tiancheng Hu, Yiru Dong, Yihan Wu, Yangxi Xiang, Xuejun Wu, Xi Chen, Buyue Qian, Yaowei Zheng, Junhao Hu
Abstract:Recursive self-improvement (RSI) offers a route to AI systems that participate in their own continued development. Recent methods allow language agents to retain changes to their parameters, memory, skills, and code for subsequent execution and improvement. We examine how these methods can inform new research ideas through two complementary perspectives. First, we introduce a unified taxonomy with nine design dimensions organized into four groups: update targets, update organization, update schedule and scope, feedback and selection. This taxonomy places existing methods in a shared design space, where examining individual dimensions and their interactions can inspire new research ideas. Second, we examine how existing RSI methods draw on traditional deep learning, for example through optimization, regularization, data augmentation, continual learning, and adversarial learning. We develop further ideas by examining more specific learning mechanisms and drawing on evolutionary computation. We also review how existing benchmarks assess self-improvement under online, offline, and offline-to-online settings. Together, these perspectives help researchers develop new RSI research ideas from existing designs and learning mechanisms. RSI papers and resources are available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.14625 [cs.AI] |
| (or arXiv:2610.14625v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.14625 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Junhao Hu [view email]
[v1]
Sat, 10 Oct 2026 06:19:13 UTC (5,144 KB)
来源：arXiv：cs.AI · arxiv.org