HEAR: Agentic Reasoner for Enterprise Systems Using Stratified Hypergraph Ontology
A new AI system called HEAR (Hypergraph Enterprise Agentic Reasoner) achieves up to 94.7% accuracy on supply-chain root cause analysis tasks. Developed to overcome hallucinations and failures in multi-hop, n-ary reasoning over heterogeneous enterprise systems, HEAR uses a Stratified Hypergraph Ontology with a Graph Layer for provenance-aware data interfaces and a Hyperedge Layer for encoding n-ary business rules. It operates an evidence-driven reasoning loop that dynamically orchestrates ontology tools without requiring LLM retraining. The system demonstrates adaptive efficiency by using procedural hyperedges to minimize token costs while leveraging topological structures. The research is detailed in arXiv paper 2605.14259.
Key facts
- HEAR achieves up to 94.7% accuracy on supply-chain root cause analysis tasks.
- HEAR uses a Stratified Hypergraph Ontology with Graph and Hyperedge layers.
- The system operates an evidence-driven reasoning loop without LLM retraining.
- HEAR demonstrates adaptive efficiency by minimizing token costs via procedural hyperedges.
- The research is published on arXiv with identifier 2605.14259.
- HEAR addresses hallucinations and failures in multi-hop, n-ary reasoning.
- Existing paradigms like GraphRAG and NL2SQL lack semantic grounding for enterprise systems.
- HEAR dynamically orchestrates ontology tools for structured multi-hop analysis.
Entities
Institutions
- arXiv