One Retrieval to Cover Them All: Co-occurrence-Aware Knowledge Base Reorganization for Session-Level RAG

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, medium

Summary

A new method addresses the limitation of Retrieval-Augmented Generation (RAG) systems, which typically optimize for single queries, by improving session-level information retrieval. Current RAG setups cover only 41% of a user's session-level information need, failing to account for coherent episodes of related questions. The proposed approach reorganizes the knowledge base (KB) offline using co-occurrence-aware clustering and expands retrieval candidates via cluster neighborhoods during query time. Tested on WixQA, a dataset of 6,221 enterprise support articles, this method increased single-query session coverage to 58% (+17% absolute; 95% CI: [14.1, 20.4]). It also reduced retrieval calls to achieve 70% coverage by 34% and compressed the KB to 20% of its original size. These improvements were consistent across four embedding models and six functional domains, advocating for session-level coverage as the primary evaluation metric for enterprise RAG.

Key takeaway

For Machine Learning Engineers building enterprise RAG systems, you should re-evaluate your knowledge base organization and retrieval strategy. Standard single-query optimization is insufficient for user sessions. Consider implementing co-occurrence-aware clustering to reorganize your KB offline and expand retrieval candidates via cluster neighborhoods. This approach can significantly boost session coverage and reduce retrieval calls, shifting your primary evaluation metric from single-query recall to comprehensive session-level performance.

Key insights

RAG systems can achieve significantly higher session-level coverage by reorganizing knowledge bases with co-occurrence-aware clustering.

Principles

• Enterprise RAG requires session-level information coverage.
• Co-occurrence clustering enhances knowledge base organization.
• Session-level coverage is key for RAG evaluation.

Method

Reorganize knowledge bases offline using co-occurrence-aware clustering. Expand retrieval candidates at query time by exploring cluster neighborhoods.

In practice

• Apply co-occurrence clustering to existing RAG KBs.
• Integrate cluster neighborhood expansion for retrieval.
• Shift RAG evaluation to session-level metrics.

Topics

• Retrieval-Augmented Generation
• Knowledge Base Reorganization
• Session-Level Retrieval
• Co-occurrence Clustering
• Enterprise RAG
• Information Retrieval Metrics

Best for: Research Scientist, AI Architect, NLP Engineer, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• One Retrieval to Cover Them All: Co-occurrence-Aware Knowledge Base Reorganization for Session-Level RAG — Reorganizing knowledge bases with co-occurrence clustering significantly boosts session-level RAG coverage and efficiency for enterprise users.
• Grounding Your LLM: A Practical Guide to RAG for Enterprise Knowledge Bases — RAG systems combine LLMs with dynamic knowledge retrieval to provide accurate, auditable, and updatable answers for enterprise data.
• Proxy-Pointer RAG: Solving Entity and Relationship Sprawl in Large Knowledge Graphs — Proxy-Pointer RAG uses structural document context for accurate entity and relationship reconciliation in knowledge graphs.
• MKG-RAG-Bench: Benchmarking Retrieval in Multimodal Knowledge Graph-Augmented Generation — Retrieval quality is a critical bottleneck and strongly determines generation outcomes in multimodal knowledge graph RAG.
• MCompassRAG: Topic Metadata as a Semantic Compass for Paragraph-Level Retrieval — Topic metadata, selected and abstracted, can efficiently guide RAG retrieval over coarse chunks, improving precision and latency.
• How Retrieval Systems Power Modern RAG Applications — RAG combines external retrieval with LLMs to overcome knowledge cutoffs and hallucinations, providing grounded, up-to-date responses.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.