Improving Self Consistency in LLMs through Probabilistic Tokenization

Open in new window