[Submitted on 1 Dec 2025 (v1), last revised 17 Oct 2026 (this version, v5)]
Abstract:Conformal regression provides finite-sample marginal coverage, but interval efficiency depends on how uncertainty is allocated across inputs. We propose emph{Conformal Laplace-Aware Predictive Scaling} (CLAPS), which combines a learned heteroscedastic scale with last-layer Laplace uncertainty. The learned variance also weights the training features used to construct the posterior, yielding a noise-weighted support correction in the learned representation. The correction contracts as weighted feature information increases and becomes most influential under localized support imbalance. Experiments show near-nominal marginal coverage and modest efficiency improvements over the matched aleatoric baseline on real regression benchmarks.
Submission history
From: Dongseok Kim [view email]
[v1] Mon, 1 Dec 2025 07:58:21 UTC (82 KB)
[v2] Tue, 9 Dec 2025 14:56:02 UTC (83 KB)
[v3] Mon, 12 Jan 2026 18:49:06 UTC (83 KB)
[v4] Tue, 5 May 2026 16:26:41 UTC (40 KB)
[v5] Sat, 17 Oct 2026 06:08:54 UTC (493 KB)
