LLM Agent Survey

Report Abuse

Basic Information

This repository hosts a comprehensive, published survey paper on Large Language Model (LLM) based autonomous agents. It systematically reviews the construction, applications, and evaluation of LLM-driven agents and identifies essential agent components such as profile, memory, planning, and action modules. The README catalogs many representative models and projects, summarizes how capability acquisition is handled, and documents applications across natural sciences, social sciences, and engineering. It records update history, figures that compare planning methods and capability acquisition, maintainers and citation information, and instructions for contributing new papers or references. The repo aggregates papers, code pointers where available, benchmark and evaluation summaries, and an interactive, maintained table of related works to support ongoing literature tracking and community contributions.

Links

App Details

Features
A structured taxonomy and large tables that enumerate and compare LLM-based autonomous agent works by components and capabilities. Detailed tables map models to modules like profile, memory, planning, action, tools, and fine-tuning, and indicate whether code or papers are available. Sections cover applications by domain and list project names with associated papers and code when provided. The README includes figures illustrating planning approaches and capability acquisition, an update log of added works, evaluation summaries with benchmarks and subjective/objective methods, citation and contributor guidance, and pointers to a maintained interactive bibliography for deeper filtering and exploration. It highlights that the survey is the first published overview in this field and invites community contributions.
Use Cases
Researchers and practitioners can use this repo as a centralized literature review and reference hub to understand the landscape of LLM-based autonomous agents. It clarifies core architectural components and tradeoffs, provides comparative tables that help locate prior art and working codebases, and summarizes evaluation practices and benchmarks to guide experimental design. The applications listing shows cross-domain use cases and helps identify relevant projects by domain and task. Figures and taxonomy support conceptual understanding of planning and capability acquisition. The maintained interactive table and contribution instructions make it easy to keep up to date and to add missing studies, making the repository useful for survey writing, project scoping, benchmarking, and identifying implementation resources.

Please fill the required fields*