分解、诊断与组合：优化智能体流水线与执行框架
原标题：Divide, Diagnose, and Compose: Optimizing Agentic Pipelines and Harnesses
AI 摘要
论文提出 DeCAF（分解式信用分配框架），用于自动优化多智能体流水线，联合调整提示词、拓扑结构、模型选择与工具策略等耦合设计维度。该方法先按基准与种子流水线自适应分解优化维度，由独立编码智能体会话在受限编辑范围内分别实验，最后综合各维度发现并用组件级轨迹评估器定位薄弱环节。在多个智能体基准上，任务性能平均提升 2.58 个百分点，推理成本降低 4.30%，端到端延迟降低 42.45%。
正文节选
Divide, Diagnose, and Compose: Optimizing Agentic Pipelines and Harnesses Abstract Building agentic pipelines that achieve high task performance while keeping cost and latency low requires jointly optimizing heterogeneous and coupled design choices, including prompts, topology, model selection, and tool policies. Manually searching this combinatorial design space is expensive and time-consuming, while existing automated optimizers typically operate within a predefined subset of this space, modif