OSU-NLP-Group/HippoRAG

[NeurIPS'24] HippoRAG is a novel RAG framework inspired by human long-term memory that enables LLMs to continuously integrate knowledge across external documents. RAG + Knowledge Graphs + Personalized PageRank.

What it solves

HippoRAG 2 is designed to overcome the limitations of standard Retrieval-Augmented Generation (RAG) systems, specifically improving their ability to handle multi-hop retrieval (associativity) and the integration of large, complex contexts (sense-making). It aims to provide LLMs with a non-parametric continual learning capability that mirrors human long-term memory.

How it works

HippoRAG 2 implements a memory framework that recognizes and utilizes connections within new knowledge. It uses an offline indexing process to build a knowledge graph from documents, which is more resource-efficient than other graph-based RAG solutions. During the online process, it retrieves information by following these connections to improve the accuracy of complex queries.

Who it’s for

This project is for developers and researchers building advanced RAG systems who need their LLMs to perform better on complex, multi-step reasoning tasks and integrate large amounts of information without sacrificing performance on simple factual queries.

Highlights

  • Improved Associativity: Enhances multi-hop retrieval capabilities compared to standard RAG.
  • Sense-Making: Better integration of large and complex contexts.
  • Resource Efficiency: Uses fewer resources for offline indexing than alternatives like GraphRAG or RAPTOR.
  • Flexible Deployment: Supports a wide range of providers including OpenAI, Amazon Bedrock, OrcaRouter, and local vLLM deployments.
  • Vector Store Support: Compatible with Parquet files, Qdrant, ChromaDB, and Milvus.

Related

  • Project
    HKUDS/LightRAG

    A lightweight, graph-based RAG framework that combines knowledge graphs and vector embeddings to provide efficient, high-quality retrieval and global context understanding.

  • Project
    iMoonLab/Hyper-RAG

    A hypergraph-driven Retrieval-Augmented Generation (RAG) framework that reduces LLM hallucinations by capturing complex, beyond-pairwise relationships in domain-specific knowledge.

  • Project
    hhy-huang/HiRAG

    HiRAG is a Retrieval-Augmented Generation framework that uses hierarchical knowledge to improve the quality and comprehensiveness of LLM responses compared to standard RAG.

  • Project
    infiniflow/ragflow

    An open-source RAG engine that combines deep document understanding and agentic capabilities to create a high-fidelity context layer for LLMs using complex unstructured data.

  • Project
    zjunlp/LightMem

    A lightweight memory management framework for LLMs and AI agents that provides efficient storage, retrieval, and update mechanisms for long-term memory.