ARTFEED — Contemporary Art Intelligence

HyCE-RAG: Hypergraph Framework for Explainable Multi-hop QA

other · 2026-07-29

A novel framework named HyCE-RAG (Hypergraph Chain-of-Evidence Retrieval-Augmented Generation) has been introduced to enhance explainability in multi-hop question answering. Traditional RAG methods depend on the semantic similarity between queries and text segments, which neglects the structural relationships among entities and facts. While graph-based RAG incorporates graph-structured knowledge, its pairwise edges are insufficient for capturing higher-order connections. In contrast, HyCE-RAG arranges entities, relationships, and contextual evidence into hyperedges, constructs a query-aware evidence hypergraph, and facilitates confidence propagation across entity-hyperedge incidence structures. This approach employs confidence-guided evidence assembly to formulate answers accompanied by explanations, effectively addressing multi-hop QA that requires gathering evidence from various documents. The research can be found on arXiv with ID 2607.22597.

Key facts

  • HyCE-RAG stands for Hypergraph Chain-of-Evidence Retrieval-Augmented Generation
  • It is designed for explainable multi-hop question answering
  • Standard RAG relies on semantic similarity between query and text chunks
  • Graph-based RAG uses pairwise edges which are limited for higher-order associations
  • HyCE-RAG organizes entities, relations, and contextual evidence into hyperedges
  • It builds a query-aware evidence hypergraph
  • It performs confidence propagation over entity-hyperedge incidence structures
  • The paper is on arXiv with ID 2607.22597

Entities

Institutions

  • arXiv

Sources