ResearchBenchmarksLarge Model Performance Prediction on Benchmark OCR pattern shiftFollow25.18RMSESTAR24.161231.038137.91544.7919Feb 12, 2026Evaluation ResultsMethodMethodLinksRMSEMAEScore AvgSRCCKRCCMAE@3%Rank Avg%TotalSTARShift Type=Benchmark-s...Shift Type=Benchmark-side, Scenario=OCR2026.0225.1819.4122.2970.6467.6217.1651.81—CPMFShift Type=Benchmark-s...Shift Type=Benchmark-side, Scenario=OCR2026.0244.9538.4441.771.7665.6111.0849.48—PMFShift Type=Benchmark-s...Shift Type=Benchmark-side, Scenario=OCR2026.0250.6544.5547.655.7353.96.8238.82—