How can a KG paper website contribute in the era of LLMs and AI agents?

[lihanghang]

built a website about knowledge graphs and language intelligence: kg.lihanghang.top. The goal is to collect foundational and recent papers and follow developments in the field.
As large language models and AI agents evolve, I’ve been thinking: Beyond helping people discover and read papers, what value could this project provide?
Knowledge graphs can provide structured knowledge for agents and may also support retrieval, long-term memory, knowledge updates, and reasoning. I’d love to hear from researchers, developers, and learners about how this website could support these emerging uses.
Here are a few possible directions:

• Track research developments: Organize key papers, evolving approaches, and open questions around KG × LLMs, GraphRAG, and agent memory.
• Connect knowledge across papers: Link papers, methods, tasks, datasets, and code to help readers understand how different works relate.
• Support reading and research: Offer paper Q&A with source citations, method comparisons, and guided reading paths.
• Make the collection accessible to agents: Provide structured paper metadata and API or MCP interfaces so agents can retrieve, cite, and organize research.
• Encourage community contributions: Collect paper recommendations, reading notes, reproduction experiences, and practical application examples.
  A few questions to start the discussion:

1. What challenges do you face when reading KG papers or building agents that this website could help address?
2. Which direction would you prioritize? Are there other ideas worth exploring?
3. Which papers, open-source projects, or real-world examples should we include?
   Concrete use cases and different perspectives are welcome. Even a single paper recommendation or an open question could help shape the project’s direction.

[lihanghang]

To make this discussion easier to join, let's start with one concrete question:

Which one paper or open-source project would you recommend to someone exploring knowledge graphs for AI agents, and why?

One starting point is From Local to Global: A Graph RAG Approach to Query-Focused Summarization. It connects entity graphs and community summaries with answering questions across a document collection. For this website, it raises an interesting question: could graph-based retrieval help readers synthesize research across papers, and how should we evaluate it against simpler retrieval approaches?

That is a question to investigate, rather than evidence that the approach will work best for this collection.

You can reply with just:

• Paper / project: A title and link.
• Why it matters: One or two sentences about the problem it addresses.
• Optional: A limitation, reproduction experience, or resource you wish existed.

English and Chinese replies are welcome.

中文：推荐一篇你认为对 KG × Agent 有价值的论文或一个开源项目，并用一两句话说明理由即可。也欢迎分享它的局限或实际使用经验。