GraphMERT: Efficient and Scalable Distillation of Reliable Knowledge Graphs from Unstructured Data

Researchers introduce GraphMERT, an 80M-parameter AI model that efficiently extracts reliable knowledge graphs from unstructured text data. The system outperforms much larger language models like Qwen3-32B in generating factually accurate and semantically valid knowledge graphs, achieving 69.8% FActScore versus 40.2% for the baseline.

• →GraphMERT is a compact 80M-parameter model that creates high-quality knowledge graphs from unstructured text data.
• →The system achieves 69.8% FActScore compared to only 40.2% for a 32B-parameter baseline LLM on medical domain data.
• →GraphMERT addresses key limitations of neurosymbolic AI frameworks including scalability and interpretability issues.
• →The model demonstrates superior reliability by reducing hallucinated relations and improving ontology consistency.
• →This represents a breakthrough in efficient knowledge graph extraction that could enable more trustworthy AI applications.