IMEX paper targets interaction-based explanations for black-box models

A new arXiv paper introduces IMEX, an interaction-based method for explaining model predictions. The authors focus on black-box models, where users can observe inputs and outputs but not a transparent internal decision process.

The work matters because explainability remains a practical barrier for AI use in regulated or high-stakes settings. Teams often need to understand why a model produced a prediction before trusting it in a workflow.

IMEX adds to a long-running research effort to make machine learning systems more interpretable without requiring full access to model internals.