RAS: Reflection-Augmented Scaling with In-Context Learning for Executable Cypher Query Generation

This paper introduces Reflection-Augmented Scaling (RAS) for executable Cypher query generation, leveraging prior execution feedback through in-context learning. RAS reduces the Query Execution Error Rate by 41-50%, significantly outperforming Independent Scaling.