RareCP: Regime-Aware Retrieval for Efficient Conformal Prediction

2026-05-09

AI Summary

Traditional conformal prediction methods struggle with time series due to temporal dependencies, distributional drift, and heterogeneous error structures, often failing to distinguish between smooth drifts and discrete error mechanisms. This work proposes RareCP, the first approach to explicitly decouple these two error types: it models discrete mechanisms via a mixture-of-cosine-attention experts, while capturing smooth drifts through a hypernetwork-driven adaptive kernel. RareCP further incorporates sparse retrieval, similarity-weighted quantiles, and parameter-space teacher anchoring to enhance robustness. Evaluated on the GIFT-Eval benchmark, RareCP significantly outperforms existing conformal prediction methods and baseline uncertainty estimators, achieving substantially narrower and more stable prediction intervals while rigorously maintaining empirical coverage guarantees.

Abstract

Recent advances in uncertainty quantification for time series forecasting show that conformal prediction can provide reliable prediction intervals, yet standard conformal methods are often inefficient under temporal dependence, drift, and heterogeneous error behavior. Existing methods typically either update miscoverage rates over time or learn unconstrained calibration weights, without explicitly separating two central sources of nonstationarity: smoothly drifting error distributions and co-existing distinct error regimes. We introduce RareCP, a regime-aware retrieval method for adaptive conformal time series prediction. RareCP learns local calibration representations through a mixture of cosine-attention experts that each capture distinct error regimes, while a compact hypernetwork adapts the kernel parameters to track temporal drift. Given a new forecasting context, RareCP retrieves the top-k most relevant calibration examples, assigns similarity weights, and forms a weighted conformal quantile over their signed residuals, yielding asymmetric prediction intervals. The adaptive kernel is trained using a smooth interval score objective, with a parameter-space anchor to a lightweight teacher kernel to preserve stable local representations. On the GIFT-Eval benchmark, RareCP improves interval efficiency over recent conformal baselines and foundation model uncertainty estimates while maintaining empirical coverage. Ablations confirm that regime-specific experts, drift-adaptive kernels, sparse retrieval, and teacher anchoring each contribute to the final performance.

Problem

• conformal prediction
• time series forecasting
• nonstationarity
• error regimes
• temporal drift

Innovation

• conformal prediction
• regime-aware retrieval
• temporal drift adaptation
• mixture of experts
• adaptive calibration