This artifact accompanies our Euro-Par 2026 paper and provides all code, scripts, datasets, and pre-trained models needed to reproduce the experiments of the BADA framework, a Bayesian-Optimization-based active-learning approach for performance prediction across RISC-V microarchitectures. The package is distributed as a single .zip archive together with larger datasets and reduced-scale subsets for fast evaluation. A fully configured Docker environment (Ubuntu 22.04, Python 3.12, PyTorch 2.5.1+cu121, CUDA 12.1) is included to avoid dependency issues. Each experiment offers two reproduction paths: complete retraining (time-consuming) and pre-trained verification (recommended, seconds to minutes) using the provided checkpoints and pre-computed statistics. Evaluation spans three environments: the StarFive VisionFive 2 RISC-V board, and the SiFive-U74 and XiangShan-Nanhu LLVM-MCA simulators. Detailed step-by-step instructions, expected outputs, and a one-click figure-generation script are included in the README to regenerate all paper figures.