Share your thoughts, 1 month free Claude Pro on usSee more

Step Verification on VisualProcessBench (test)

71.32Macro-F1 (DynaMath)

Gemini 2.5 Pro

Updated 16d ago

Evaluation Results

Method
Gemini 2.5 Pro 2026.03	71.32	64.42	69.78	65.26	72.43	68.64	-
EVPV-PRM 2026.03	69.57	68.86	67.09	65.27	69.11	67.46	-
doubao-seed-1.6-vision 2026.03	68.66	61.86	65.57	62.51	64.62	64.91	-
Gemini 2.5 Pro 2026.03	68.47	63.34	68.26	65.15	69.48	67.13	-
Qwen3-VL-235B-A22B-instruct 2026.03	68.43	61.54	65.97	64.54	64.16	65.45	-
doubao-seed-1.6-vision 2026.03	66.19	59.47	63.12	61.07	62.74	62.77	-
TIM-PRM-8B 2026.03	65.9	58.3	61.9	58.3	63.9	61.7	-
BETAPRM 2026.05	63.75	59.49	62.91	62.19	66.69	-	62.91
BETAPRM 2026.05	63	60.48	60.7	60.43	59.98	-	60.97
Standard PRM 2026.05	62.93	63.12	61.93	59.9	63.79	-	61.9
Standard PRM 2026.05	62.91	62.05	60.47	60.78	64.38	-	61.54
Standard PRM 2026.05	62.88	61.44	61.25	62.17	65.55	-	62.23
BETAPRM 2026.05	62.71	62.42	62.17	59.65	64.23	-	61.85
VisualPRM-8B 2026.03	62.7	58.5	61	62.1	61.8	62	-
Qwen3-VL-30B-A3B-instruct 2026.03	62.27	59	59.68	56.49	59.5	59.26	-
BETAPRM 2026.05	61.67	59.18	62.74	60.94	64.59	-	61.9
Standard PRM 2026.05	61.63	58.73	60.41	60.2	63.23	-	60.69
qwen2.5-vl-72b-instruct 2026.03	61.43	60.25	59.85	59.12	59.72	59.99	-
Qwen3-VL-30B-A3B-instruct 2026.03	58.95	61.29	57.37	57.49	58.76	58.22	-
QWEN-VL-PRM-7B 2026.03	58.3	55.8	58.8	55.7	59.8	58.6	-
gpt-4o-mini 2026.03	58.13	53.2	54.09	52.07	54.62	54.29	-
Qwen3-VL-235B-A22B-instruct 2026.03	57.63	58.73	58.08	59.59	58.76	58.51	-
qwen2.5-vl-72b-instruct 2026.03	56.99	59.43	56.43	58.09	55.72	57.19	-
gpt-4o-mini 2026.03	56.57	54.08	52.53	51.42	56.74	53.57	-
Base (w/o training) 2026.05	54.85	50.21	52.04	52.4	49.95	-	52.28
Base (w/o training) 2026.05	50.19	48.86	50.97	47.61	47.34	-	49.4
Base (w/o training) 2026.05	49.62	49.85	49.58	50.22	48.51	-	49.68
Base (w/o training) 2026.05	49.41	47.01	48.91	47.21	49	-	48.39