Large language models have sparked an explosion of research into multi-agent systems powered by these models — but the field is chaotic and rapidly evolving. The taichengguo/LLM_MultiAgents_Survey_Papers repository serves as a curated, academic-grade guide through this complexity. It collects and organizes the growing body of literature into a structured taxonomy that helps researchers and practitioners alike understand the key streams of progress and open challenges.
What the LLM multi-agent survey repository provides
This repository is not a software framework or runnable codebase. Instead, it acts as a living bibliography and research companion for the IJCAI 2024 paper titled “Large Language Model based Multi-Agents: A Survey of Progress and Challenges.” It tracks papers, datasets, and benchmarks related to multi-agent systems that harness large language models (LLMs).
The core of the repo is a categorization of the fast-growing research area into five distinct streams:
Frameworks: General-purpose multi-agent platforms and toolkits designed to build and orchestrate LLM agents, exemplified by projects like CAMEL, AutoGen, and MetaGPT.
Orchestration and efficiency: Research focusing on how to coordinate multiple agents effectively, ensuring scalability, efficient communication, and resource usage.
Problem solving: Practical applications of multi-agent collaboration for tasks such as software development, embodied robotics, and scientific discovery.
World simulation: Using multiple agents to model and simulate complex environments or scenarios.
Datasets and benchmarks: Collections of tasks, challenges, and datasets designed to evaluate multi-agent LLM systems.
The repository is updated bi-weekly to include the latest relevant publications, capturing the evolution from early works around 2023 to papers projected as far forward as 2026. Alongside the bibliographic data, it provides synthesized reference architecture diagrams illustrating how multi-agent LLM systems are generally structured according to the surveyed literature.
The project’s main content is organized as markdown files, each summarizing papers and categorizing them by their research focus. This makes it straightforward for researchers to scan, filter, and pinpoint papers relevant to their interests.
How the taxonomy clarifies a fragmented research field
The technical strength of this repository lies in its carefully crafted taxonomy and comprehensive coverage. Multi-agent LLM research is growing so fast that the landscape can be overwhelming and fragmented. By distilling the field into five clear categories, the repo reveals two emerging camps:
General-purpose agent frameworks: Platforms like AutoGen, CAMEL, and MetaGPT build reusable, extensible agent orchestration layers aimed at a broad range of tasks. They focus on agent role definition, communication protocols, and system efficiency.
Domain-specific multi-agent applications: Other research channels apply multi-agent collaboration to concrete domains such as code generation, embodied robotics, and scientific modeling. These works prioritize domain efficacy and task-specific optimizations over framework generality.
This distinction highlights a central tension in the field — the tradeoff between building flexible, scalable frameworks versus optimizing for domain-specific problem solving. The repository’s curated bibliography and reference architecture help researchers navigate this tension, understand where their work fits, and identify gaps or opportunities.
While the repo itself does not provide executable code or benchmarking results, its curated nature and frequent updates make it a valuable compass in a rapidly evolving area. It reduces the noise and duplication by organizing hundreds of papers, many from 2024 and beyond, into a digestible map.
Explore the project
The repository is structured primarily as a set of markdown documents and bibliographic data files. The README outlines the survey paper’s abstract and the taxonomy rationale. Key markdown files group papers by category and provide brief summaries and links to original sources.
The repo also includes diagrams illustrating reference architectures for LLM-based multi-agent systems, synthesizing insights from the surveyed literature. These visual artifacts are worth a close look if you want to understand typical system components and agent interactions.
Because this is a research companion rather than a software library, there are no installation or runtime commands. The best way to use the project is to explore the categorized markdown files, track new updates every two weeks, and use it as a launching point for deeper dives into specific research threads.
Verdict
For researchers, AI practitioners, and developers interested in multi-agent systems powered by large language models, this repository is a valuable reference resource. It consolidates and organizes a sprawling, fast-moving body of work into a coherent map that helps you keep up with new developments and situate your own research.
The repo’s limitations are clear: it is not a framework or toolkit you can deploy or build on directly. Instead, it serves as a curated academic bibliography and survey companion, updated frequently to reflect the latest papers and trends.
If you are tracking progress in LLM-based multi-agent collaboration or planning research in this area, the taxonomy and reference architectures here are worth understanding. They reveal the fundamental architectural tradeoffs and show the split between general-purpose frameworks and application-driven research. For anyone building or evaluating multi-agent LLM systems, this repository helps you see the forest for the trees.
Related Articles
- Mapping the LLM agent landscape with the awesome-llm-agents curated catalog — A curated catalog of 20+ LLM agent frameworks and tools organized by agent type and capabilities. Understand architectur
- llm-wiki: orchestrating multi-agent LLM research into persistent knowledge bases — llm-wiki is a shell-based orchestration layer that turns LLM agents into a persistent, multi-agent research wiki. Suppor
- A practical taxonomy for large language model ensembles: Exploring the Awesome-LLM-Ensemble repository — The Awesome-LLM-Ensemble repo catalogs research on combining multiple LLMs with a clear three-phase taxonomy: before, du
- Awesome LLM Apps: a practical collection of runnable AI agent and RAG templates — Awesome LLM Apps offers 100+ runnable AI agent and RAG templates for quick LLM app development. It supports multiple pro
- A hands-on course for mastering large language models: fine-tuning, quantization, and tooling — Explore a comprehensive LLM course with practical notebooks on fine-tuning (QLoRA, DPO), quantization (GPTQ), and tools
→ GitHub Repo: taichengguo/LLM_MultiAgents_Survey_Papers ⭐ 1,265