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

Multimodal Reasoning on VisualProcessBench Overall

54.6FEI Average

VRPRM-MiMo

Updated 12d ago

Evaluation Results

Method	Links
VRPRM-MiMo 2025.08		54.6	72.91
VRPRM-MiMo w/o CoT 2025.08		53.03	71.62
VRPRM-Qwen3 2025.08		51.82	70.01
VRPRM-MiMo w/o RL & w/o CoT 2025.08		51.61	46.77
MiMo-VL-7B 2025.08		51.09	63.38
VRPRM-MiMo w/o RL 2025.08		50.77	60.61
VRPRM 2025.08		49.72	66
VRPRM-Qwen3 w/o RL 2025.08		49.24	57.07
Qwen3-VL-4B-Thinking 2025.08		49.18	61.4
Qwen2.5-VL-7B 2025.08		48.37	45.36
VRPRM-Qwen3 w/o RL & w/o CoT 2025.08		48.3	53.39
VRPRM-Qwen3 w/o CoT 2025.08		47.14	68.63
Qwen2.5-VL-72B 2025.08		46.54	47.82
VRPRM w/o RL 2025.08		45.77	62.11
VRPRM w/o RL & w/o CoT 2025.08		44.58	52.2
InternVL2.5-8B 2025.08		43.89	46.08
Gemini-2.0-Flash 2025.08		43.68	42.17
VRPRM w/o CoT 2025.08		38.14	53.66
GPT-4o-mini 2025.08		36.45	35.91
VisualPRM-8B 2025.08		25.61	61.03