让智能体自动设计并优化规划结构，提升复杂任务适应力。
TodoEvolve: Learning to Architect Agent Planning Systems
- 用模块化设计空间统一多种规划方式，支持灵活组合
- 在5个基准上优于人工设计的规划模块，且资源开销低
- 适合需要自适应规划能力的研究者与开发者
规划已成为现代智能体系统应对复杂、长周期任务的核心能力，但现有方法多依赖固定的手工设计规划结构，难以适应开放问题的多样性。为此，我们提出 TodoEvolve，一种元规划范式，可自主合成并动态调整特定任务的规划架构。首先构建 PlanFactory，一个模块化设计空间，将拓扑、初始化、自适应和导航等不同规划范式标准化为统一代码库，提供异构规划模式的通用接口。基于 PlanFactory 收集高质量规划轨迹，并通过 extit{Impedance-Guided Preference Optimization}（IGPO）训练 Todo-14B，这是一种多目标强化学习目标，鼓励生成在任意任务和智能体底座下兼具高性能、稳定性与令牌效率的规划系统。在五个智能体基准上的实证评估表明，TodoEvolve 均显著优于精心设计的规划模块，同时保持较低的 API 成本和运行时开销。
原文摘要 · Abstract (English)
Planning has become a central capability for contemporary agent systems in navigating complex, long-horizon tasks, yet existing approaches predominantly rely on fixed, hand-crafted planning structures that lack the flexibility to adapt to the structural diversity of open-ended problems. To address this limitation, we introduce TodoEvolve, a meta-planning paradigm that autonomously synthesizes and dynamically revises task-specific planning architectures. Specifically, we first construct PlanFactory, a modular design space that standardizes diverse planning paradigms within a unified codebase encompassing topology, initialization, adaptation, and navigation, thereby providing a common interface for heterogeneous planning patterns. Leveraging PlanFactory, we collect high-quality planning trajectories and train Todo-14B via \textit{Impedance-Guided Preference Optimization} (IGPO), a multi-objective reinforcement learning objective that encourages the generation of planning systems that are performant, stable, and token-efficient across arbitrary tasks and agent backbones. Empirical evaluations on five agentic benchmarks demonstrate that TodoEvolve consistently surpasses carefully engineered planning modules while maintaining economical API costs and runtime overhead.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。