文章摘要
2026年六大AI顶会共收录165篇Agent方向相关论文,Graph Agent是该领域突出的新兴研究方向。针对当前领域资源分散、学习路径不清晰的问题,有整理者梳理出含355篇论文、821个项目的分级学习资源集,用户添加指定微信回复关键词“Agent”即可领取对应资源站链接。

From the 2026 six major top conferences' publication statistics, research papers related to the Agent direction show remarkable scale. Among them, AAAI contributes 57 papers, CVPR 47, ICLR 43, and ICML, ECCV, IJCAI add up to 18 papers, making the total number of Agent-related papers reach 165.

Among various subfields, Graph Agent stands out as a most prominent new direction. AAAI 2026 has released several high-quality works in this area, including S-DAG, MPAS, DoM, and HAGG.

While the field is hot and offers abundant opportunities, a practical problem arises: relevant resources are too scattered. Summaries, classic papers, latest top conference works, and open-source projects are spread across different platforms, leading to cluttered collections and unclear learning paths.

Recently, a systematic collation of Agent-related papers and projects was completed, forming a usable resource set. The collection includes 355 papers and 821 projects, organized into six sections following a learning logic from beginner to topic selection.

The six sections of the resource are as follows:

  • Review papers (13): Start with these to quickly build a global understanding of the field.
  • Core system forms (34): Focus on two main lines: single-agent and multi-agent systems.
  • Technical foundations and integration innovations (38): Cover areas like Agent combined with reinforcement learning, Agent Memory, and Graph+Agent.
  • Learning support resources (45): Includes basic content such as Transformer, LoRA, CoT, and RAG for supplementary learning.
  • Project resources: Contains 321 industry cases and 500 open-source projects from code hosting platforms.
  • Top conference papers (165): Covers the six 2026 conferences mentioned earlier.

Additionally, key classic content such as ReAct, Reflexion, Voyager, AutoGen, MemGPT, and DeepSeek-R1 has been separately categorized, eliminating the need to search across multiple platforms.

To obtain the curated resource, you can add a small assistant's WeChat account, reply with the keyword Agent, and the link will be sent automatically without extra steps.

If you know colleagues or peers interested in Agent research or planning to switch to this direction, feel free to share this information. Helping someone save a month of exploration time is a valuable contribution.

以上内容不代表本平台立场,仅供读者参考