ResearchBenchmarksMathematical Reasoning on GSM8k (Accuracy, Perplexity)Follow86.7AccuracyMAGS85.45285.77686.186.424May 20, 2026Evaluation ResultsMethodMethodLinksAccuracyPerplexityMAGSBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0586.71.194UnsteeredBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.05861.196CDBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0585.81.271ASBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0585.71.335ITIBase Model=Llama-3.1-8...Base Model=Llama-3.1-8B-Instruct2026.0585.51.335