Large language model (LLM) agents are fundamentally constrained by context length on long-horizon tasks. We introduce Context-Folding, a framework that empowers agents to actively manage their working context. An agent can procedurally branch into a sub-trajectory to handle a subtask and then fold it upon completion, collapsing the intermediate steps while retaining a concise summary of the outcome. To make this behavior learnable, we develop an end-to-end reinforcement learning framework FoldGRPO with specific process rewards to encourage effective task decomposition and context management. On complex long-horizon tasks (Deep Research and SWE), our folding agent matches or outperforms the ReAct baselines while using an active context 10× smaller and significantly outperforms models that rely on summarization-based context management.
@article{arxiv.2510.11967,
title = {Scaling Long-Horizon LLM Agent via Context-Folding},
author = {Weiwei Sun and Miao Lu and Zhan Ling and Kang Liu and Xuesong Yao and Yiming Yang and Jiecao Chen},
journal= {arXiv preprint arXiv:2510.11967},
year = {2025}
}