Conformal Inference
Conformal inference is a model-agnostic framework for constructing prediction sets that guarantee a user-specified probability of containing the true value, regardless of the underlying data distribution. Current research focuses on extending conformal methods to handle non-exchangeable data (like time series), adapting to distribution shifts, and improving efficiency through techniques like adaptive scoring and optimized algorithms (e.g., using regression trees or flow-based models). This robust approach to uncertainty quantification is increasingly important for reliable decision-making in diverse applications, including medical imaging, personalized medicine, and safety-critical systems, where understanding prediction uncertainty is paramount.