Query The Quantum

Short summary

A TigerGraph hackathon project benchmarks GraphRAG against basic RAG and LLM-only approaches on 2M+ quantum computing papers, demonstrating that knowledge graph retrieval achieves 90% factual accuracy with 60% fewer tokens consumed. Graph-based multi-hop reasoning outperforms vector similarity on complex queries linking quantum algorithms, hardware, and research relationships. Reproducible code on GitHub uses free-tier TigerGraph, Groq, and open-source tools.

• •GraphRAG achieves 90% LLM-as-Judge accuracy vs 80% (LLM-only) and 70% (basic RAG) on quantum computing queries
• •Demonstrates 60% token reduction and 20-30% latency improvement compared to vector-based RAG
• •Fully reproducible benchmark with code, setup instructions, and free-tier API configurations on GitHub

Generated with AI, which can make mistakes.