HippoRAG: Hippocampal Memory of RAG

Priciples, Examples, and Code

Many important real-world tasks, including scientific literature review, legal case briefing, and medical diagnosis, require knowledge integration across passages or documents.

Existing RAG methods are unable to aid LLMs in tasks that demand the integration of new information across passage boundaries, as each new passage is encoded independently.

HippoRAG is a novel retrieval framework, taking inspiration from hippocampal indexing theory of human long-term memory. It collaborates with LLMs, knowledge graphs, and personalized PageRank algorithms. This collaboration mimics the varied roles of the neocortex and hippocampus in human memory.