LLM 기반 자율 에이전트에 대한 서베이
아키텍처, 협업, 응용
본 서베이는 LLM 기반 자율 에이전트의 아키텍처와 협업 패턴을 검토하며, 프로파일링, 메모리, 계획, 행동 모듈을 다룬다.
A Survey on Large Language Model-Based Agents: Architecture, Memory, and Multi-Agent Collaboration
Authors: [Your Name] Date: March 13, 2026 Status: Draft v1.0
Large Language Models (LLMs) have emerged as a foundational component for building autonomous agents capable of perceiving environments, making decisions, and executing complex tasks. This survey provides a comprehensive review of LLM-based agents, covering their architectural design, memory mechanisms, and multi-agent collaboration paradigms. We systematically analyze 15 representative works published between 2023 and 2025, categorizing agent architectures into profiling, planning, and action modules. We further examine memory systems ranging from in-context short-term memory to vector database-backed long-term memory, and explore emerging multi-agent frameworks that enable cooperative problem-solving. Finally, we discuss domain-specific applications in software engineering, finance, and gaming, alongside evaluation challenges and future research directions.
Keywords: Large Language Models, Autonomous Agents, Multi-Agent Systems, Memory Mechanism, Survey