Oliver Prompts says PageIndex scores 98.7% on FinanceBench without vector search

Oliver Prompts says researchers have developed PageIndex, a retrieval-augmented generation approach that does not use a vector database, data embeddings, chunking, or similarity search. Instead of splitting documents into chunks and loading them into Pinecone, it builds a tree index so a language model can work through the documents.

Prompts says PageIndex scored 98.7% on FinanceBench and outperformed every vector RAG system on that leaderboard. He also says it is fully free and open source.