One Retrieval to Cover Them All: Co-occurrence-Aware Knowledge Base Reorganization for Session-Level RAG
arXiv:2606.31156v1 Announce Type: cross Abstract: RAG systems retrieve documents optimized for answering one query at a time. Yet enterprise users arrive with sessions, that is, coherent episodes of related questions that span semantically distant parts of the knowledge base. We show that a single retrieval call over a standard knowledge base covers only 41% of a user's session-level information need. To close this gap, we reorganize the KB offline using co-occurrence-aware clustering and expand retrieval candidates through cluster neighborhoods at query time. On WixQA (6,221 enterprise suppor
The proliferation of RAG systems highlights their limitations for complex, multi-query user interactions, making advancements in session-level knowledge retrieval critical for enterprise adoption and efficiency.
Improving RAG systems to handle session-level information needs significantly enhances the utility and efficiency of AI for enterprise users, reducing the need for fragmented, single-query interactions and enabling more sophisticated AI agentic workflows.
RAG systems are evolving beyond single-query optimization to incorporate 'co-occurrence-aware knowledge base reorganization,' allowing them to address broader user sessions and semantically distant information needs more effectively.
- · Enterprise AI users
- · AI software developers
- · Customer support sectors
- · Knowledge management platforms
- · Inefficient RAG systems
- · Traditional keyword-based retrieval
Enterprises will see an increase in the effectiveness and accuracy of their AI-powered knowledge retrieval systems, leading to better decision-making and operational efficiency.
The improved ability of RAG to handle complex, multi-topic user sessions could accelerate the development and deployment of more capable AI agents for white-collar workflows.
As AI systems become more adept at understanding and navigating vast, semantically diverse knowledge bases, the foundational requirements for human-in-the-loop oversight in knowledge work might shift towards higher-level strategic guidance rather than micro-management.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI