When Punctuation Matters: A Large-Scale Comparison of Prompt Robustness Methods for LLMs

Mikhail Seleznyov, Mikhail Chaichuk, Gleb Ershov +3

Large Language Models (LLMs) are highly sensitive to subtle, non-semantic variations in prompt phrasing and formatting. In this work, we present the first systematic evaluation of…