AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

CLEAR: Calibrated Learning for Epistemic and Aleatoric Risk

Ilia Azizi, Juraj Bodik, Jakob Heiss, Bin Yu

ICLR 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Epistemic versus aleatoric decomposition

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
combine aleatoric and epistemic uncertainty with two calibrated parameters (CLEAR) instead of PCS alonePCS ensemble · 17 regression datasetsquantile loss: 15.8% improvementImproved
add epistemic uncertainty and tune lambda on top of residual-based aleatoric estimate (ablation)ALEATORIC-R · 17 regression datasetsquantile loss: 9.4%Improved
fit aleatoric quantiles on residuals instead of directly on targets (ablation)bootstrapped CQR · 17 regression datasetsquantile loss: 34.4% -> 9.4% gap to CLEARImproved
replace UACQR with CLEARUACQR-S / UACQR-P · 17 regression datasetsNCIW, AISL, width: CLEAR outperforms both on 14 of 17 datasetsImproved
replace UACQR with CLEAR on insuranceUACQR-S / UACQR-P · insuranceNCIW: -15.5% (vs UACQR-S), -18.6% (vs UACQR-P)Hurt
apply CLEAR to Deep EnsemblesDeep Ensembles · 17 regression datasetsNCIW: +28.57%Improved
apply CLEAR to SQRSQR · 17 regression datasetsNCIW: +13.36%Improved
apply CLEAR to conformally calibrated SQRSQR-conformal · 17 regression datasetsPICP: -0.15%Hurt
learn both gamma1 and lambda instead of fixing lambda=1 or gamma1=1 (ablation)CLEAR · 17 regression datasetsinterval qualityMixed
use CLEAR instead of PCS or CQR with 2 featuresAmes Housing · Ames Housing, 2 featuresquantile loss: 3,818 -> 3,131Improved
use CLEAR instead of PCS with all featuresAmes Housing · Ames Housing, all featuresquantile loss: 1,922 -> 1,923No clear effect
reduce training data to 20% so lambda grows (ablation)CLEAR · Ames Housing, all features, 20% datalambda: 14.45 -> 100Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy