arXiv：cs.AI· Zixuan Xie, Chang Xu, Yuhua Jiang, Li Zhao·· 16 小时前
HyperAgents 新方法：用可编辑 Improvement Dynamics Model 预测自我进化智能体的改进效果
Predict Before You Modify: Learning Improvement Dynamics for Self-Evolving Agents
AI 导读
研究者提出可编辑的 Improvement Dynamics Model（IDM），在评估前预测修改的任务级结果并指导可选修订，评估后用历史修改与结果持续改进 IDM。在 HyperAgents 基础上，该方法在多语言编程、论文评审、机器人奖励设计和奥数级数学评分四个领域均优于 HyperAgents，平均性能最高提升 46%，预测知识还可迁移到新领域。
正文
Abstract:Self-evolving agents improve through repeated cycles of modification and evaluation. However, existing methods receive feedback on modifications only after costly evaluations, and aggregate scores may obscure distinctions between failures that require different repairs. We introduce predictive self-evolution through an editable Improvement Dynamics Model (IDM). Before evaluation, the IDM predicts task-level outcomes of a proposed modification and guides an optional revision. After evaluation, past modifications and their observed outcomes are used to improve the IDM, allowing predictive knowledge to accumulate across generations. The system can thus improve how it solves tasks, proposes modifications, and predicts their effects. Building on HyperAgents, we evaluate this approach across multilingual coding, paper review, robotics reward design, and Olympiad-level mathematical grading. Our approach outperforms HyperAgents across all four domains, with up to 46% higher average performance over generations. Gains also persist under matched token budgets that include both agent evolution and task evaluation, and extend to held-out tasks. Forecast accuracy improves as the IDM evolves, and the learned predictive knowledge transfers to improve evolution in a new domain.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.13991 [cs.LG] |
| (or arXiv:2610.13991v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13991 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zixuan Xie [view email]
[v1]
Sat, 26 Sep 2026 18:59:47 UTC (2,245 KB)
来源：arXiv：cs.AI · arxiv.org