Taming Hyperparameter Sensitivity in Data Attribution: Practical Selection Without Costly Retraining | Read Paper on Bytez