ResearchBenchmarksMathematical Reasoning on GSM8K (EM Strict/Flex)Follow38.89EM (Strict)Baseline36.841237.373137.90538.4369May 26, 2026Evaluation ResultsMethodMethodLinksEM (Strict)EM (Flex)BaselineEvaluation Framework=l...Evaluation Framework=lm-eval-harness2026.0538.8939.04ReMoEEvaluation Framework=l...Evaluation Framework=lm-eval-harness2026.0538.1338.36CE-onlyEvaluation Framework=l...Evaluation Framework=lm-eval-harness, Training=Generic router fine-tuning2026.0536.9237.23