LZ Penalty: An information-theoretic repetition penalty for autoregressive language models

Open in new window