InfiAgent: An Infinite-Horizon Framework for General-Purpose Autonomous Agents

2026-01-06

arXiv.org

AI Summary

This work addresses the limitations of large language model (LLM) agents in long-horizon tasks, where unbounded context growth and error accumulation hinder performance. Existing approaches struggle to balance information fidelity with reasoning stability. To overcome this, the authors propose InfiAgent, a novel framework that introduces an explicit state externalization mechanism. It abstracts persistent state into a file-centric workspace and combines periodic state snapshots with a fixed-size window of recent actions, enabling bounded reasoning context independent of task length. This design transcends the constraints of conventional context-centric paradigms. Evaluated on DeepResearch and an 80-paper literature review task, InfiAgent achieves performance comparable to larger closed-source systems without fine-tuning and significantly outperforms baseline methods in long-horizon task coverage.

Technology Category

• Multiagent Systems: Modeling other Agents
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

LLM agents can reason and use tools, but they often break down on long-horizon tasks due to unbounded context growth and accumulated errors. Common remedies such as context compression or retrieval-augmented prompting introduce trade-offs between information fidelity and reasoning stability. We present InfiAgent, a general-purpose framework that keeps the agent's reasoning context strictly bounded regardless of task duration by externalizing persistent state into a file-centric state abstraction. At each step, the agent reconstructs context from a workspace state snapshot plus a fixed window of recent actions. Experiments on DeepResearch and an 80-paper literature review task show that, without task-specific fine-tuning, InfiAgent with a 20B open-source model is competitive with larger proprietary systems and maintains substantially higher long-horizon coverage than context-centric baselines. These results support explicit state externalization as a practical foundation for stable long-horizon agents. Github Repo:https://github.com/ChenglinPoly/infiAgent

Problem

• long-horizon tasks
• context growth
• error accumulation
• reasoning stability
• autonomous agents

Innovation

• Infinite-horizon agents
• State externalization
• File-centric state abstraction
• Long-horizon reasoning
• Context-bounded LLM agents