ResearchBenchmarksMathematical Reasoning on GSM8K DE (8-shot test)Follow47.5AccuracyT-Free41.2642.8844.546.12Mar 16, 2026Evaluation ResultsMethodMethodLinksAccuracyBytes per Sequence PositionT-Freeshots=8-shot, evaluati...shots=8-shot, evaluation_mode=Pre-training2026.0347.54.38HATifiedshots=8-shot, evaluati...shots=8-shot, evaluation_mode=Pre-training2026.0344.14.37Llamashots=8-shot, evaluati...shots=8-shot, evaluation_mode=Pre-training2026.0341.52.94