Enhancing Contextual Understanding in Large Language Models through Contrastive Decoding

Zheng Zhao, Emilio Monti, Jens Lehmann +1

Large language models (LLMs) tend to inadequately integrate input context during text generation, relying excessively on encoded prior knowledge in model parameters, potentially re…