siRBench

Repository layout (quick map):

• benchmark/: Docker-based competitor pipeline (prepare/train/test wrappers). Outputs go to benchmark/competitors/updated_validation_results/ by default, or benchmark/competitors/original_results/ when running with --original.
• data/: CSV splits, feature columns, and helper scripts (train/val/test/leftout).
• siRBench-model/: Standalone model (XGBoost + LightGBM) with uv setup and training/inference scripts.