Repetition penalty
Repetition penalty is a technique used in natural language processing, particularly in language generation models, to reduce repetitive outputs. By penalizing the likelihood of previously generated tokens, it encourages more diverse and coherent text generation.