negative sampling - A training technique that uses negative (non-target) samples to help a model distinguish true data from noise.