Improving the Efficiency of LLM Agent Systems through Trajectory Reduction

2025-09-27

AI Summary

To address the high input-token overhead caused by ever-growing reasoning trajectories in large language model (LLM)-based multi-turn agents for software engineering tasks, this paper proposes AgentDiet—a training-free, inference-time trajectory compression method. We first systematically identify pervasive redundancy, irrelevance, and obsolescence in agent trajectories; leveraging this insight, we design a lightweight, dynamic filtering strategy that automatically prunes context while preserving task-critical information, ensuring compatibility across diverse LLMs and software engineering tasks. Experiments on two mainstream LLMs and standard benchmarks demonstrate that AgentDiet reduces input tokens by 39.9%–59.7% and computational cost by 21.1%–35.9%, with no degradation in agent performance. Our core contributions are threefold: (i) the first quantitative characterization of trajectory redundancy in LLM-based agents; (ii) a novel, efficient, general-purpose, and training-free trajectory compression paradigm; and (iii) empirical validation of its effectiveness and robustness across models and tasks.

Technology Category

• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Multi-turn agent systems based on Large Language Models (LLMs) have been increasingly popular for software engineering tasks. While LLM agents show decent effectiveness, the high computational cost of input tokens due to the ever-growing trajectory remains an efficiency concern for their applications. Efficiency is largely neglected in existing studies and agent products, and this paper fills the gap by introducing an inference-time trajectory reduction approach to reduce the cost of agents. Through analyzing existing agent trajectories, we demonstrate that useless, redundant, and expired information is widespread in all trajectories, which can be identified and reduced without harming the agent's performance. We then design a simple yet effective trajectory reduction approach, AgentDiet, which automatically removes such waste information. We implement AgentDiet on a top-performing coding agent, and the evaluation on two LLMs and two benchmarks shows that AgentDiet can reduce input tokens by 39.9% ~ 59.7%, or the final computational cost by 21.1% ~ 35.9%, while maintaining the same agent performance. This indicates that trajectory reduction is a promising direction for agent systems.

Problem

• Reducing computational costs in LLM agent systems
• Identifying and removing redundant trajectory information
• Maintaining agent performance while improving efficiency

Innovation

• Introduces inference-time trajectory reduction approach
• Automatically removes useless redundant expired information
• Reduces input tokens by up to 59.7% without performance loss