Hierarchical Path-Addressed Retrieval (HPAR)
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Current approaches to grounding AI agents in external knowledge---flat Retrieval-Augmented Generation (RAG), knowledge graphs, and static skill files---share a limitation: they separate retrieved content from its organizational context. A chunk of text is returned, but where it lives, what surrounds it, and what contains it is lost. This paper proposes Hierarchical Path-Addressed Retrieval (HPAR): a grounding approach built on an infinitely nestable, human-seeded, AI-maintained outline tree, where every node is addressable by its full path, and where that path itself serves as a contextual signal for retrieval. We argue that HPAR offers a combination of properties---bidirectional traversal, structural locality, pointer-native data references, emergent naming, importance-ordered siblings, and continuous co-refinement---that are difficult or expensive to achieve in existing approaches, and that these properties together constitute a promising foundation for AI agent grounding. No empirical evaluation is presented; this paper is an architectural proposal and an invitation to investigation.
Fields of Study & Subjects
Information Retrieval and Search Behavior
Topic Modeling
AI-based Problem Solving and Planning
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 28 March 2026
- Language
- EN