Tag
A new benchmark, FML-Bench, reveals that recent improvements in MLE-Bench scores are largely due to better base models and increased search budget rather than algorithmic advances.