KAG

lastmod 2025-01-14created 2025-01-141 edits

KAG is a logical form-guided reasoning and retrieval framework based on OpenSPG engine and LLMs. It is used to build logical reasoning and factual Q&A solutions for professional domain knowledge bases. It can effectively overcome the shortcomings of the traditional RAG vector similarity calculation model.

KAG - Knowledge Graph RAG Framework

Features

• A logical form-guided reasoning and Q&A framework based on the OpenSPG engine and LLMs
  • Designed to build logical reasoning and Q&A solutions over vertical domain knowledge bases
• Overcomes the limitations of existing RAG:
  • Ambiguity of vector similarity calculation in traditional RAG
  • Solves the noise problem of OpenIE-based GraphRAG
• Supports logical reasoning and multi-hop Q&A
• Outperforms recent SOTA methods
• Fully integrates the logical/factual characteristics of knowledge graphs (KGs)
• Core features:
  • Integrates complete contextual information through a mutual indexing structure between knowledge and chunks
  • Mitigates the noise problem through knowledge alignment via conceptual semantic reasoning
  • Supports representation of domain expert knowledge through schema-constrained knowledge construction
  • Supports logical/multi-hop Q&A through logical form-based hybrid reasoning

See also

• OpenSPG
• LLM

Favorite site

• Github - OpenSPG/KAG