MARL-Papers is a curated paper collection for multi-agent reinforcement learning research.
The repository addresses the challenge of navigating the expanding landscape of multi-agent reinforcement learning literature by organizing papers chronologically across multiple research dimensions. It serves researchers seeking to understand the field's foundations and current directions by providing a structured reference list that connects MARL to related areas including single-agent reinforcement learning, game theory, evolutionary computation, and applications in large language models and robotics.
Researchers new to multi-agent reinforcement learning should use this collection as a starting point for understanding the field's scope and evolution. The papers are organized by topic areas including joint action learning, cooperation and competition, coordination, security, self-play, learning to communicate, transfer learning, imitation learning, meta-learning, and networked MARL with decentralized training and execution. The collection also includes dedicated sections for MARL applications in large language models and robotics, making it particularly valuable for those exploring these emerging application domains. Tutorials and foundational books are listed separately to help newcomers build conceptual understanding before diving into specialized research papers.
The project welcomes community contributions through pull requests and maintains an open approach to paper inclusion, with a stated principle that authors may request removal of their work. The repository is organized as a straightforward reference list rather than a code implementation or framework, making it accessible to anyone seeking to survey the research landscape without requiring technical setup or dependencies.
