arXiv：cs.AI· Xin Xu, Ke Cheng, Hao Wang, Yangru Huang, Yanghao Xiao, Zhichao Chen, Zhisheng Zhong, Xun Liu, Yi Li, Qian Li, Haibin Lv, Jun Zhang, Jie Jiang, Zhouchen Lin·· 19 小时前
SpecRSI：面向智能体系统的推测式递归自我改进
Drafting Agents, Not Just Tokens: Speculative Recursive Self-Improvement for Agentic Systems
AI 导读
SpecRSI 让可执行的后继智能体在被完全评估前就先行起草，但只有通过完整验证才会被采纳。其配套的 VATS 分配编辑与待验证检查的预算，GVO 则从已完成搜索中学习改进器。在主要科学对比中，其终端增益超过 Janus 33.3%–40.0%，并在全部 50 个科学主干—任务组合上优于 Janus；编程验证成功率达 69.4%，高于 HGM 的 64.3%，目标成本指数低 20.8%。
正文
Authors:Xin Xu, Ke Cheng, Hao Wang, Yangru Huang, Yanghao Xiao, Zhichao Chen, Zhisheng Zhong, Xun Liu, Yi Li, Qian Li, Haibin Lv, Jun Zhang, Jie Jiang, Zhouchen Lin
Abstract:Improving an agent through successive revisions requires balancing exploration and evaluation under a limited budget. A narrow search can miss useful sequences of edits, while fully evaluating every intermediate version can exhaust the budget before those sequences yield a verified improvement. We introduce Speculative Recursive Self-Improvement (SpecRSI), which drafts executable successor agents before their predecessors are fully evaluated, while requiring complete verification before adoption. Verification-Aware Tree Search (VATS) allocates the budget between further edits and outstanding checks using the predicted verified return of a continuation. Group Verification Optimization (GVO) learns the improver from completed searches while keeping the real evaluator fixed. In the main scientific comparison, terminal gains exceed Janus by 33.3%--40.0%. Additional evaluations show that SpecRSI outperforms Janus in all 50 scientific backbone--task pairs and exceeds MetaRSI-v1 by 1.85%--2.38% in macro-average system score across twenty backbones. Coding-validation success reaches 69.4% versus HGM's 64.3% at a 20.8% lower target-cost index.
| Comments: | 47 pages; code available at this https URL |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.16491 [cs.LG] |
| (or arXiv:2610.16491v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.16491 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xin Xu [view email]
[v1]
Mon, 12 Oct 2026 09:43:48 UTC (1,743 KB)
来源：arXiv：cs.AI · arxiv.org