An Open Benchmark for Testing RAG on Realistic Company-Internal Data
EnterpriseRAG-Bench released a 500k-document corpus for testing RAG on company-internal data. It simulates Redwood Inference across 9 sources and includes 500 questions over 10 retrieval failure modes. Baselines show BM25 beats vector search overall, while agentic/bash retrieval has the best completeness at higher cost and latency.
Why it matters: EnterpriseRAG-Bench tests internal-company RAG with 500k documents and 500 questions across 9 data sources and 10 retrieval failure modes; BM25 beat vector retrieval overall.