Mathematical Reasoning Process Evaluation on PROCESSBENCH

82.9GSM8K Accuracy

Llama-3.3-70B-Instruct

Updated 4d ago

Evaluation Results

Method
Llama-3.3-70B-Instruct 2025.12	82.9	59.4	46.7	43	58
GPT-4o-0806 2025.12	79.2	63.6	51.4	53.5	61.9
Qwen2.5-72B-Instruct 2025.12	76.2	61.8	54.6	52.2	61.2
Llama-3.1-70B-Instruct 2025.12	74.9	48.2	46.7	41	52.7
Step Consistency 2025.12	72	67.9	64.5	65.7	67.5
Skywork-PRM-7B 2025.12	70.8	53.6	22.9	21	42.1
Meta Critic + Outcome Consistency 2025.12	70.2	68.7	63.4	65.2	66.9
Reference Guided 2025.12	70.2	67.4	63.7	64.1	66.4
Meta Critic + Outcome Consistency 2025.12	70	67.1	59.2	58.3	63.7
Outcome Consistency 2025.12	69.5	68.7	61.4	60.5	65
Qwen2.5-14B-Instruct 2025.12	69.3	53.3	45	41.3	52.2
Outcome Consistency 2025.12	68.3	68	66.6	63.3	66.6
Qwen2.5-Math-7B-PRM800K 2025.12	68.2	62.6	50.7	44.3	56.5
Meta Critic 2025.12	68	67.2	62.4	65.5	65.8
Step Consistency 2025.12	67.6	70.3	63.9	61	65.7
Single Verification 2025.12	67	65.5	61.2	62	63.9
Reference Guided 2025.12	66.1	66.6	60.2	59.8	63.2
Qwen2.5-Math-72B-Instruct 2025.12	65.8	52.1	32.5	31.7	45.5
Qwen2.5-32B-Instruct 2025.12	65.6	53.1	40	38.3	49.3
Meta Critic 2025.12	65.3	66	60.4	56.6	62.1
Single Verification 2025.12	62.5	63	56.8	57	59.8
Skywork-PRM-1.5B 2025.12	59	48	19.3	19.2	36.4
RLHFlow-PRM-Mistral-8B 2025.12	50.4	33.4	13.8	15.8	28.4
Math-Shepherd-PRM-7B 2025.12	47.9	29.5	24.8	23.8	31.5
RLHFlow-PRM-Deepseek-8B 2025.12	38.8	33.8	16.9	16.9	26.6
Qwen2.5-7B-Instruct 2025.12	36.5	36.6	29.7	27.4	32.6
Qwen2.5-Math-7B-Instruct 2025.12	26.8	25.7	14.2	12.7	19.9
Llama-3.1-8B-Instruct 2025.12	10.9	5.1	2.8	1.6	5.1