Revisiting ORM Performance Anti-Patterns in LLM-Generated Data Access Code: Replication Artifact

Replication artifact for the paper "Revisiting ORM Performance Anti-Patterns in LLM-Generated Data Access Code", accepted at the 2026 IEEE 18th International Scientific Conference on Informatics. The artifact contains experimental configurations, task specifications, generation prompts, ORM scaffolding, runtime measurements, anti-pattern detection rules, human validation data, and analysis scripts. The study evaluates 900 generation attempts across three language models, three ORM ecosystems (Java/Hibernate, PHP/Doctrine, and Node.js/Prisma), five data-access tasks, and two prompt conditions. The provided scripts support reproduction of the reported aggregate results, tables, and figures from the archived measurements. The artifact does not guarantee identical outputs from a new execution of the language models.