PIPE-Cypher: Automatic Enterprise Benchmark Generation for Text-to-Cypher Systems
Researchers have developed PIPE-Cypher, an automated pipeline for generating benchmarks for text-to-Cypher systems. This system addresses the challenge of creating relevant benchmarks for enterprise property graphs, which often have unique schemas and evolving structures. PIPE-Cypher uses a local LLM to generate diverse and executable query pairs, ensuring they reflect real-world usage and can be validated against the graph. AI
IMPACT Enables more accurate and adaptable evaluation of text-to-Cypher models in enterprise settings.