The Inference Capabilities of Large Language Models May Be Overestimated: Significant Weaknesses in Unfamiliar Scenarios This translated title succinctly captures the essence of the original Chinese statement, highlighting the concern that the reasoning abilities of large language models might be overly praised, particularly when faced with unfamiliar situations where they exhibit notable shortcomings. The title maintains a neutral tone, presenting a critical perspective on the current understanding of these models' capabilities.