Toward Efficient Agents: A Survey of Memory, Tool learning, and Planning

Xiaofang Yang1,2,† Lijun Li1,† Heng Zhou1,3,† Tong Zhu1,† Xiaoye Qu1 Yuchen Fan1,4 Qianshan Wei5 Rui Ye4 Li Kang1,4 Yiran Qin6 Zhiqiang Kou7 Daizong Liu8 Qi Li5 Ning Ding9 Siheng Chen4 Jing Shao1
1Shanghai Artificial Intelligence Laboratory, 2Fudan University,
3University of Science and Technology of China, 4Shanghai Jiaotong University,
5Institute of Automation, Chinese Academy of Sciences,
6The Chinese University of Hong Kong (Shenzhen),
7Hong Kong Polytechnic University, 8Wuhan University, 9Tsinghua University
† Indicates Main contributors
Paper Code arXiv

Abstract

Recent years have witnessed increasing interest in extending large language models into agentic systems. While the effectiveness of agents has continued to improve, efficiency, which is crucial for real-world deployment, has often been overlooked. This paper therefore investigates efficiency from three core components of agents: memory, tool learning, and planning, considering costs such as latency, tokens, steps, etc. Aimed to conducting comprehensive research addressing the efficiency of the agentic system itself, we review a broad range of recent approaches that differ in implementation yet frequently converge on shared high-level principles including but not limited to bounded context via compression and retrieval, reduced action cost via budgeted tool use and caching, and controlled search via hierarchical planning and pruning for improving efficiency, which we discuss in detail. Accordingly, we characterize efficiency in two complementary ways: comparing effectiveness under a fixed cost budget, and comparing cost at a comparable level of effectiveness. This trade-off can also be viewed through the Pareto frontier between effectiveness and cost. From this perspective, we also examine efficiency oriented benchmarks by summarizing evaluation protocols for these components and consolidating commonly reported efficiency metrics from both benchmark and methodological studies. Moreover, we discuss the key challenges and future directions, with the goal of providing promising insights.

Paper List Navigation

📑 Table of Memory Contents

Memory Architecture Overview

In the paper, we organize memory into construction, management, and access. Since many papers overlap across these stages, this part is primarily organized around memory construction to avoid redundancy.

🔄 Working Memory

💾 External Memory

📦 Item-based Memory

👥 Multi-Agent Memory

📑 Table of Tool Learning Contents

Tool Learning Architecture Overview

Tool Learning framework encompasses tool selection, tool calling, and tool-integrated reasoning for enhanced agent capabilities.

🔧 Tool Selection

🏷️ Multi-Label Classification (MLC)

▶️ Tool Calling

📝 In-Place Parameter Filling

🎯 Efficient Tool Calling with Post-training

🛠️ Tool-Integrated Reasoning (TIR)

📑 Table of Planning Contents

Planning Architecture Overview

Planning framework encompasses single-agent planning efficiency and multi-agent collaborative strategies for enhanced decision-making.

🤖 Single-Agent Planning Efficiency

👥 Multi-Agent Collaborative Efficiency

BibTeX

@misc{yang2026efficientagentsmemorytool,
      title={Toward Efficient Agents: Memory, Tool learning, and Planning}, 
      author={Xiaofang Yang and Lijun Li and Heng Zhou and Tong Zhu and Xiaoye Qu and Yuchen Fan and Qianshan Wei and Rui Ye and Li Kang and Yiran Qin and Zhiqiang Kou and Daizong Liu and Qi Li and Ning Ding and Siheng Chen and Jing Shao},
      year={2026},
      eprint={2601.14192},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2601.14192}, 
}