Understanding the nuanced performance of machine learning models is essential for responsible deployment, especially in high-stakes domains like healthcare and finance. This paper introduces a novel framework, Conformalized Exceptional Model Mining, which combines the rigor of Conformal Prediction with the explanatory power of Exceptional Model Mining (EMM). The proposed framework identifies cohesive subgroups within data where model performance deviates exceptionally, highlighting regions of both high confidence and high uncertainty. We develop a new model class, mSMoPE (multiplex Soft Model Performance Evaluation), which quantifies uncertainty through conformal prediction's rigorous coverage guarantees. By defining a new quality measure, Relative Average Uncertainty Loss (RAUL), our framework isolates subgroups with exceptional performance patterns in multi-class classification and regression tasks. Experimental results across diverse datasets demonstrate the framework's effectiveness in uncovering interpretable subgroups that provide critical insights into model behavior. This work lays the groundwork for enhancing model interpretability and reliability, advancing the state-of-the-art in explainable AI and uncertainty quantification.

Music rattled the windows of the one-room schoolhouse that was now serving as a dance floor for nearly the entire village, a population of about 100 people. Masato, a masticated yuca drink, was passed around the room. I tried to refuse it as it came to me -- I had already shared an entire pot and was feeling woozy from both the alcohol and my full stomach. But this was a celebration and another bowl was pressed into my hands. The party was the last night of my first field trip to the Amazon in 2012.