Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligence
A curated collection of papers, benchmarks, and open-source projects on graph engineering for coordinating LLM agents, organized around a survey moving from individual agent intelligence to system-level intelligence.
This repository accompanies a survey on graph engineering for LLM agents, collecting research papers, benchmarks, and open-source projects that use explicit, dynamic graph structures to organize tasks, coordinate heterogeneous agents, maintain runtime state, and support system evolution. It follows the survey's progression across three levels: model intelligence, built through pretraining, post-training, prompt engineering, and context engineering; individual intelligence, where a single agent gains tools, memory, skills, and persistent interaction loops; and system intelligence, where graph and ontology engineering organize multiple agents, tasks, and runtime state into a coordinated whole.
The repository is organized as a table of contents linking to related survey papers and a body of research papers grouped under these levels, and is continuously updated as new relevant work appears; contributions are welcome by issue or pull request.
It accompanies an arXiv preprint and is licensed under MIT.