A field guide to the LLM multi-agent paper zoo

What it does This repository is the companion bibliography to an IJCAI 2024 survey on LLM-based multi-agent systems. The authors collect and classify papers into five streams—frameworks, orchestration, problem solving, world simulation, and benchmarks—so you can find relevant work without drowning in arXiv alerts. It functions as a living literature map that the authors aim to update roughly every two weeks.

The interesting bit The taxonomy is ambitious enough to span everything from multi-robot navigation and protein discovery to simulated economies and disease propagation, suggesting the authors view this less as a narrow subfield and more as a general computing pattern. The included architecture diagram and overview table attempt to distill the whole sprawl into a single mental model.

Key highlights

• Anchored to a peer-reviewed survey (IJCAI 2024) rather than being an unofficial list.
• Covers applied domains often ignored in pure-ML reading lists: embodied agents, scientific teams, database administration, and social simulation.
• Explicitly maintained with a stated goal of biweekly updates and community contributions.
• Groups papers by functional purpose (framework vs. orchestration vs. simulation) rather than by model size or training technique.

Caveats

• The repository is a bibliography with paper links; there is no code, reproduction scripts, or evaluation harness.
• The README source provided is truncated, so some sections and entries are incomplete in the current view.

Verdict Grab this if you are writing a literature review or trying to find the one relevant paper in a suddenly crowded field; skip it if you are hunting for reusable libraries or training pipelines.