ResearchBenchmarksMathematical Reasoning on MATH first 1,000 samples (test)Follow86.8Exact Match AccuracyMGRS77.9680.25582.5584.845Nov 28, 2025Evaluation ResultsMethodMethodLinksExact Match AccuracyMGRSBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1186.8AoTBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1183.6AFlowBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1183FoTBackbone=GPT-4o-mini, n=8Backbone=GPT-4o-mini, n=82025.1182.5CoT-SCBackbone=GPT-4o-mini, n=5Backbone=GPT-4o-mini, n=52025.1181.8Self-refineBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1178.7CoTBackbone=GPT-4o-miniBackbone=GPT-4o-mini2025.1178.3