All publications

arXiv 2026 · Preprint

ARC-STAR: Auditable Post-Hoc Correction for PDE Foundation Models

Chengze Li, Lingwei Wei, Li Sun, Hongbo Lv, Jie Yang, Hanrong Zhang, Kening Zheng, Wei-Chieh Huang, Enze Ma, Philip S. Yu

Overview

ARC-STAR corrects forecasts from a frozen PDE foundation model without fine-tuning the underlying solver. It first addresses global bias, then refines local residual errors, and allocates limited computation to regions identified as high risk. Separating these stages makes their contributions measurable and supports correction under a compute budget.

Resources