K2RAG combines knowledge graph retrieval, hybrid dense/sparse search, summarization, and a quantized LLM to achieve slightly higher answer similarity (mean 0.57) than naive RAG baselines on MultiHopRAG with lower VRAM and faster training.
Fabbri, Caiming Xiong, and Chien-Sheng Wu
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
KeyKnowledgeRAG (K^2RAG): An Enhanced RAG method for improved LLM question-answering capabilities
K2RAG combines knowledge graph retrieval, hybrid dense/sparse search, summarization, and a quantized LLM to achieve slightly higher answer similarity (mean 0.57) than naive RAG baselines on MultiHopRAG with lower VRAM and faster training.