RareCP: Regime-Aware Retrieval for Efficient Conformal Prediction

TL;DR

RareCP introduces a regime-aware retrieval method for adaptive conformal time series prediction, effectively handling nonstationarity and error regimes to produce reliable and efficient prediction intervals.

Contribution

It presents a novel mixture of cosine-attention experts and a hypernetwork to explicitly model error regimes and temporal drift in conformal prediction for time series.

Findings

Improves interval efficiency over recent conformal baselines.

Maintains empirical coverage while adapting to nonstationarity.

Ablation studies confirm the effectiveness of regime-specific experts and drift-adaptive kernels.

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.