Beyond Predicting Responses: Conformal Inference for Latent Distributional Parameters

We develop LatentCP, a conformal framework that transfers calibrated uncertainty in observable response space back to an unobserved latent parameter through a known forward model. The method provides finite-sample marginal coverage without latent calibration labels, a unique inverse mapping, or knowledge of the latent mixing distribution.

Recommended citation: Minxing Zheng, Wenbin Zhou, and Shixiang Zhu. (2026). "Beyond Predicting Responses: Conformal Inference for Latent Distributional Parameters." arXiv:2608.03607.
Download Paper