TodoEvolve: Learning to Architect Agent Planning Systems

TL;DR

TodoEvolve is a novel meta-planning framework that autonomously synthesizes adaptable planning architectures for agents, outperforming fixed structures across diverse benchmarks with efficient resource use.

Contribution

It introduces PlanFactory for modular planning design and employs IGPO for training adaptable, high-performing planning systems.

Findings

Outperforms fixed planning modules on five benchmarks

Maintains low API costs and runtime overhead

Generates adaptable planning architectures

Abstract

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.