Wikipedia August 2025 (MoNaCo subset) - Neo4j - Harrier v1 0.6b embedded

This record contains the pre-built knowledge graph used in "Reducing Hallucinations in Complex Question Answering using Simple Graph-based Retrieval-Augmented Generation (Wedge, Stutter, Dixon, and Cała, arXiv:2606.05901): a Neo4j-ready Parquet export of the curated English Wikipedia subset over which the paper's agentic RAG system was evaluated, covering the source pages of all answerable questions within the MoNaCo benchmark. The graph was built from the 2025-08-01 English Wikipedia multistream dump. Articles are decomposed into Article -> Section -> Paragraph -> Chunk nodes, where Chunk nodes are 500-character slices of paragraph text with 40-character overlap. Chunk text and Article titles carry embeddings from microsoft/harrier-oss-v1-0.6b. If you use this dataset, please cite the paper above.