ResearchBenchmarksLoss-driven active learning on GP synthetic functions in-distributionFollow0.09Weighted RMSEPOLAR0.07760.16130.2450.3287Jun 23, 2026Evaluation ResultsMethodMethodLinksWeighted RMSEWeighted NLPDPOLARbackbone=TabICLv2, lay...backbone=TabICLv2, layer=last layer2026.060.092.07GP-VRω2026.060.111.65GP-US2026.060.131.26GP-VR2026.060.171.18GP-RS2026.060.40.18