Researchers introduced VESPO, a new mathematical method that corrects imbalances in trial-and-error AI training to keep the process stable and boost performance.