The field of machine learning is witnessing significant developments in performative prediction and conformal prediction. Researchers are working to relax assumptions and generalize models to nonlinear cases, preserving essential theoretical properties. Notable advancements include the development of algorithms that guarantee performative stability and efficient full conformal prediction frameworks. These innovations have the potential to improve the reliability and trustworthiness of AI systems.

Some noteworthy papers in this area include: Nonlinear Performative Prediction, which presents a novel design that generalizes performative prediction to nonlinear cases while preserving essential theoretical properties. Optimizing In-Context Learning for Efficient Full Conformal Prediction, which introduces an efficient FCP framework that employs a permutation-invariant Transformer-based ICL model trained with a CP-aware loss. Towards Performatively Stable Equilibria in Decision-Dependent Games for Arbitrary Data Distribution Maps, which proposes a gradient-based sensitivity measure that directly quantifies the impact of decision-induced distribution shifts. CloudFormer: An Attention-based Performance Prediction for Public Clouds with Unknown Workload, which proposes a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. Beyond the Pre-Service Horizon: Infusing In-Service Behavior for Improved Financial Risk Forecasting, which proposes a novel framework aimed at improving pre-service risk prediction through the integration of in-service user behavior data.