ResearchBenchmarksMathematical Reasoning on MATH 500 (Accuracy, Perplexity)Follow53AccuracyMAGS47.59248.99650.451.804May 20, 2026Evaluation ResultsMethodMethodLinksAccuracyPerplexityMAGSBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.05531.227ASBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0550.61.232ITIBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0549.81.227CDBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0549.21.295UnsteeredBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0547.81.229