Benchmarks	MMLU-Redux	GPQA			MMLU-Pro
Default Setting	0-shot CoT JSON	0-shot CoT			5-shot CoT
Models		Main (448)	Diamond (198)	Extended (546)
ArmoRM-Llama3-8B-v0.1	54.23	33.3	33.3	31.3
Meta-Llama-3-8B-Instruct	61.66
Meta-Llama-3.1-8B-Instruct	67.24	32.8			48.3
Meta-Llama-3-70B-Instruct	78.01
Meta-Llama-3.1-70B-Instruct	82.97	46.7			66.4