Text this: Unraveling overoptimism and publication bias in ML-driven science - Data for Experiment 2 and 3