ResearchBenchmarksModel Ranking Correlation Analysis on SEED-Bench (10%-50% subsets)Follow0.94Spearman Correlation (10% subset)M2IRT0.39920.53960.680.8204Mar 3, 2026Evaluation ResultsMethodMethodLinksSpearman Correlation (10% subset)Spearman Correlation (20% subset)Spearman Correlation (30% subset)Spearman Correlation (40% subset)Spearman Correlation (50% subset)M2IRT2026.030.940.960.970.970.95M3IRT2026.030.940.950.950.950.95FlashEval2026.030.890.90.90.90.9IRT2026.030.860.830.840.810.79Random2026.030.860.870.890.910.91TinyBenchmarks2026.030.690.750.770.80.81MIRT2026.030.420.590.710.770.82