RAGU: a multi-step GraphRAG engine with a compact domain-adapted LLM

Novosibirsk State University / ITMO University

Research official 2 src. ~1 min

RAGU is an open-source GraphRAG system that separates entity extraction from consolidation via a six-stage pipeline (typed extraction, DBSCAN deduplication, community detection), paired with Meno-Lite-0.1, a 7B fine-tuned extraction model that outperforms Qwen2.5-32B on knowledge-graph construction by 12.5%.

Why it matters

Shows that a compact, domain-adapted 7B model can beat a much larger general model on GraphRAG construction while running on a single GPU, arguing that extraction/comprehension skills scale more slowly with model size than factual knowledge does.

Importance: 2/5

Notable systems contribution but a niche application (GraphRAG tooling) rather than a broad capability jump.

rag graph-rag small-models knowledge-graphs

Sources

official RAGU on HF Daily Papers

official RAGU arXiv preprint