Abstract

Recent advances in large language models (LLMs) have achieved remarkable results across a wide range of natural language processing (NLP) applications. As LLMs grow increasingly capable, the need to control their generation outcome becomes more pressing,...