Mathematical Reasoning Process Evaluation on PROCESSBENCH

82.9GSM8K AccuracyLlama-3.3-70B-Instruct

Evaluation Results

Method
Llama-3.3-70B-Instruct2025.12	82.9	59.4	46.7	43	58
GPT-4o-08062025.12	79.2	63.6	51.4	53.5	61.9
Qwen2.5-72B-Instruct2025.12	76.2	61.8	54.6	52.2	61.2
Llama-3.1-70B-Instruct2025.12	74.9	48.2	46.7	41	52.7
Step Consistency2025.12	72	67.9	64.5	65.7	67.5
Skywork-PRM-7B2025.12	70.8	53.6	22.9	21	42.1
Meta Critic + Outcome Consistency2025.12	70.2	68.7	63.4	65.2	66.9
Reference Guided2025.12	70.2	67.4	63.7	64.1	66.4
Meta Critic + Outcome Consistency2025.12	70	67.1	59.2	58.3	63.7
Outcome Consistency2025.12	69.5	68.7	61.4	60.5	65
Qwen2.5-14B-Instruct2025.12	69.3	53.3	45	41.3	52.2
Outcome Consistency2025.12	68.3	68	66.6	63.3	66.6
Qwen2.5-Math-7B-PRM800K2025.12	68.2	62.6	50.7	44.3	56.5
Meta Critic2025.12	68	67.2	62.4	65.5	65.8
Step Consistency2025.12	67.6	70.3	63.9	61	65.7
Single Verification2025.12	67	65.5	61.2	62	63.9
Reference Guided2025.12	66.1	66.6	60.2	59.8	63.2
Qwen2.5-Math-72B-Instruct2025.12	65.8	52.1	32.5	31.7	45.5
Qwen2.5-32B-Instruct2025.12	65.6	53.1	40	38.3	49.3
Meta Critic2025.12	65.3	66	60.4	56.6	62.1
Single Verification2025.12	62.5	63	56.8	57	59.8
Skywork-PRM-1.5B2025.12	59	48	19.3	19.2	36.4
RLHFlow-PRM-Mistral-8B2025.12	50.4	33.4	13.8	15.8	28.4
Math-Shepherd-PRM-7B2025.12	47.9	29.5	24.8	23.8	31.5
RLHFlow-PRM-Deepseek-8B2025.12	38.8	33.8	16.9	16.9	26.6
Qwen2.5-7B-Instruct2025.12	36.5	36.6	29.7	27.4	32.6
Qwen2.5-Math-7B-Instruct2025.12	26.8	25.7	14.2	12.7	19.9
Llama-3.1-8B-Instruct2025.12	10.9	5.1	2.8	1.6	5.1