arXiv cs.AI· Huan Zhang, Mingju Chen, Dongxu Zhou, Can Lv, Sen Cui, Faguo Wu, Shiji Zhou·· 2 天前AI 评分35
RSIFlow：带进化元智能体的自适应自我改进工作流
RSIFlow: Adaptive Self-Improvement Workflows with an Evolving Meta-Agent
AI 导读
RSIFlow 是一个递归自我改进框架，将自适应工作流编排与元智能体能力进化相结合，通过任务-智能体轨迹和环境反馈诊断性能瓶颈，动态选择组件级改进工作流。元智能体将工作流结果蒸馏为组件级技能与跨组件原则，进化自身 harness 以优化后续编排。在六个基准和两个骨干模型上，RSIFlow 在所有基准上取得 SOTA 平均性能，消融实验验证了自适应编排与元 harness 进化的贡献，代码已开源。
正文 · 原文
Abstract:Large language model (LLM) agents have demonstrated strong capabilities in complex tasks, including code generation, tool use, and multi-hop question answering. Existing self-improvement methods enhance agent performance through model optimization, harness evolution, and executable artifact refinement, while recursive self-improvement (RSI) further extends optimization to the improvement mechanisms themselves. However, adapting self-improvement workflows to evolving performance bottlenecks while improving workflow orchestration through execution feedback remains a key challenge. We propose \textbf{RSIFlow}, a \textbf{R}ecursive \textbf{S}elf-\textbf{I}mprovement framework that integrates adaptive work\textbf{flow} orchestration with meta-agent capability evolution. RSIFlow diagnoses performance bottlenecks from task-agent trajectories and environmental feedback to dynamically select component-specific improvement workflows. The meta-agent then distills workflow outcomes into component-specific skills and cross-component principles, evolving its own harness to improve subsequent workflow orchestration. Experiments across six benchmarks and two backbone models show that RSIFlow achieves the SOTA average performance on all benchmarks. Ablation studies further demonstrate the contributions of adaptive workflow orchestration and meta-harness evolution. Code is available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.21791 [cs.AI] |
| (or arXiv:2610.21791v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21791 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Huan Zhang [view email]
[v1]
Thu, 15 Oct 2026 12:45:08 UTC (1,888 KB)
来源：arXiv cs.AI · arxiv.org