Toward Efficient Agents: Memory, Tool learning, and Planning

TL;DR

The paper investigates efficiency in LLM-based agents by dissecting memory, tool learning, and planning as the key contributors to cost-effective, real-world deployment. It formalizes an efficiency framework, highlighting how memory construction, management, and access, along with selective tool use and bounded planning, shape latency, token usage, and compute costs, while introducing a Pareto-frontier perspective on performance versus resource expenditure. It surveys a wide range of methods—textual, latent, external, graph-based, and hierarchical memories; tool-selection and calling strategies; and single- and multi-agent planning paradigms—grounding them in concrete benchmarks and metrics. The work emphasizes practical deployment considerations, such as online vs offline memory management, parallel tool calling, cost-aware planning, and coordination in multi-agent systems, and it identifies gaps in standardized efficiency evaluation and latent reasoning for agents. Overall, the survey provides a structured blueprint for designing, evaluating, and comparing efficient agent systems, with guidance toward unified benchmarks and deployment-aware architectures that balance task success with resource consumption.

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 at 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 bounding context via compression and management, designing reinforcement learning rewards to minimize tool invocation, and employing controlled search mechanisms to enhance 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.

Figures (5)

• Figure 1: The evolutionary trajectory of efficient agent research. The diagram is organized into four principal branches: Memory, Tool Learning, Planning, and Benchmarks. Key works and their institutional affiliations are mapped chronologically to illustrate the field's development and categorization from 2023 to 2025.
• Figure 2: From LLMs to agents: standalone reasoning to trajectory-level reasoning with memory, planning, and tool learning, while introducing additional cost sources.
• Figure 3: Efficient memory overview. This figure summarizes the agent-memory lifecycle in three phases: Memory Construction, which compresses long interaction context in working and external memory to mitigate token explosion; Memory Management, which curates and updates an accumulating memory store via rule-based, LLM-based, or hybrid strategies to control latency; and Memory Access, which determines what memories to retrieve and how to integrate them into the model.
• Figure 4: Efficient tool learning comprises three stages: Tool Selection identifies candidate tools via retrieval or classification; Tool Calling handles parameter filling and execution with a focus on cost-aware constraints and budget feedback and Tool-Integrated Reasoning optimizes efficient reasoning trajectories through selective invocation and policy optimization.
• Figure 5: Overview of Efficient Planning. It aims to maximize task success while minimizing costs. (a) Single-agent methods optimize inference strategies (control, search, decomposition) or evolve via learning (policy, memory). (b) Multi-agent methods reduce overhead via topological optimization, context optimization, and coordination distillation.