arXiv cs.CL· Bangrui Xu, Jie Yang, Sidi Chen, Xuanhe Zhou, Jiantao Qiu, Qianyu Yang, Zijie Wang, Fan Wu, Zhoufutu Wen·· 11 小时前AI 评分31
SelfImprove-Bench：评估自我改进智能体的泛化能力
SelfImprove-Bench: Evaluating Generalization in Self-Improving Agents
AI 导读
研究者提出 SelfImprove-Bench，一个针对递归自我改进（RSI）系统的评估框架，将任务组织为探索、规划、推理三类能力场景，并设置 G0 到 G4 的泛化阶梯等级衡量能力增益的迁移程度。实验显示，现有 RSI 方法在不同泛化等级上的进化增益并不一致。
正文 · 原文
Abstract:Recursively self-improving (RSI) systems are increasingly viewed as a promising means of accelerating progress toward artificial general intelligence. They aim to autonomously improve their capabilities by refining their working pipelines and recursively improving the strategies that guide subsequent refinements, with gains generalizing to unseen this http URL, existing evaluations lack reliable metrics to distinguish capability gains from scenario fitting, sustainable and extensible protocols for unseen tasks and difficulties, and fine-grained attribution of improvement to evolution traces. To this end, we propose SelfImprove-Bench,an RSI evaluation framework that organizes tasks into three primary capabilities (i.e., Exploring, Planning, Reasoning) as scenarios for guiding model evolution. We instantiated the tasks on games for controllability and reproducibility, and pairs them with scalable generation and orchestration driven by task templates and difficulty coefficients. Beyond task-level scores, SelfImprove-Bench organizes evaluation into generalization ladder levels G0 to G4 that measure how far gains transfer from training conditions. Experiments reveal that the evolutionary gains of existing RSI methods are inconsistent across generalization levels.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2611.02797 [cs.AI] |
| (or arXiv:2611.02797v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.02797 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Bangrui Xu [view email]
[v1]
Mon, 2 Nov 2026 09:08:12 UTC (556 KB)
来源：arXiv cs.CL · arxiv.org