<ul><li>HippoRAG is a paper that introduces a neurobiologically inspired long-term memory for large language models.</li><li>HippoRAG combines a knowledge graph (KG) and Personalized PageRank (PPR) to achieve a single-step multi-hop retrieval process, making it faster, more accurate, and cheaper compared to traditional RAG solutions.</li><li>The knowledge graph is built by leveraging large language models for Open Information Extraction (OpenIE) and retrieval encoders for linking entities.</li><li>HippoRAG also uses PageRank to order the results based on real-world relevance.</li></ul>

About HippoRAG

Discover more