LLM-Compiled Wikis Outperform Vector RAG for Educational Corpus Grounding

A comparative study shows that synthesizing a corpus into linked concept pages via an LLM-compiled wiki provides superior contextual grounding compared to standard vector-based RAG. The approach focuses on ingest-time knowledge representation rather than query-time retrieval optimization.