355 Recent Agent Papers + 821 Projects: A Free, Organized Resource

In 2026, six major conferences accepted a total of 165 Agent papers (AAAI 57, CVPR 47, ICLR 43, ICML 7, ECCV 6, IJCAI 5), a dramatic rise from two years earlier when Agent work was scattered across many topics.

The author compiled these 355 papers and 821 project implementations into a public repository, preserving original links and organizing them by research dimensions.

The repository is divided into six sections:

Survey papers : 13 surveys, including the collaborative "Toward Efficient Agents" from nine institutions that covers memory, tool learning, and planning.

By agent count : 21 single‑agent papers and 13 multi‑agent papers, detailing design of individual agents and coordination among many agents.

Technical fusion : 4 large‑model agent papers, 2 graph‑agent papers, 17 reinforcement‑learning agent papers, and 15 agent memory‑system papers, each grouped by sub‑topic.

Learning support resources : 45 items ranging from foundational works such as "Attention Is All You Need" to recent advances in multimodal learning, inference, safety alignment, and Agentic RL.

Project resources : 821 real‑world cases, including 321 projects curated by Google and 500 open‑source agents spanning health, finance, education, and retail, useful for related‑work citations.

2026 conference breakdown : detailed counts per conference (AAAI 57, CVPR 47, ICLR 43, ICML 7, ECCV 6, IJCAI 5).

Reading guidance recommends first studying 2–3 surveys to build an overall map, then deep‑reading five papers in the chosen sub‑area (single or multi‑agent), followed by weekly scans of the relevant conference sections to stay up‑to‑date and spark new research ideas.

The entire resource is freely accessible, aiming to save researchers months of searching and organizing effort.