AI Agents Papers
A biweekly-updated, curated collection of research papers on AI agent applications and architecture, organized by capability, architecture, operations and application.
AI Agents Papers curates recent research on the applications and architectural techniques of AI agents, systems that perceive an environment, reason through tasks, and use tools to act toward a goal. The maintainers run weekly arXiv searches with specific keywords and add a paper only when it introduces a distinctively new approach, rather than aiming for exhaustive coverage.
Entries are filed into four layers: capabilities such as reasoning, planning and memory; architecture, covering how single- and multi-agent systems are built; operations, covering how agents are run and evaluated for safety; and applications, covering where agents are used. Recent additions and high-activity topics are marked with badges, and a taxonomy file documents the rules for where each paper is filed.