Assessing and Enhancing the Robustness of Large Language Models with Task Structure Variations for Logical Reasoning

Qiming Bao, Gael Gendron, Alex Yuxuan Peng +5

Large language models (LLMs), such as LLaMA, Alpaca, Vicuna, GPT-3.5 and GPT-4, have advanced the performance of AI systems on various natural language processing tasks to human-li…