A fork of the official JetBrains Long Code Arena baselines, extended with new models, metrics, retrieval strategies, and prompt-engineering experiments across three benchmarks related to coding tasks.