Improving Instruction Following in Language Models through Proxy-Based Uncertainty Estimation

JoonHo Lee, Jae Oh Woo, Juree Seok +11

Assessing response quality to instructions in language models is vital but challenging due to the complexity of human language across different contexts. This complexity often resu…