Evaluating and Explaining Prompt Sensitivity of LLMs Using Interactions
Published in ICML, 2026
Recommended citation: Qin, R., Wang, Q., Wang, T., Wei, Z., & Shen, W. Evaluating and Explaining Prompt Sensitivity of LLMs Using Interactions. In ICML 2026. https://icml.cc/virtual/2026/poster/65089
Abstract. This paper introduces an interaction-based metric to evaluate and explain prompt sensitivity in large language models from a fine-grained perspective.
Authors: Ruiyang Qin, Qingzhuo Wang, Tian Wang, Zhihua Wei, Wen Shen.