In natural language processing, transformer-based large language models
(LLMs) like GPT-x models developed by OpenAI have revolutionized the landscape.
Despite their impressive capabilities, these models often encounter challenges
when handling tasks that differ from their training dat
本研究发现 Large Language Models 的 Chain-of-Thought Reasoning (思维的串联过程)能够提供合理的解释,但有时会受到输入偏见的影响而误导我们对模型预测的真实原因的认识,这可能导致人们过度信任 LLMs 的预测结果,因此有必要针对模型解释的忠实度进行有针对性的评估和改进,特别是在社会偏见问题方面。