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

Multimodal Reasoning on MathVerse Vision Only

58.76Accuracy

InternVL2.5-38B + VRPRM

Updated 12d ago

Evaluation Results

Method	Links
InternVL2.5-38B + VRPRM 2025.08		58.76
InternVL2.5-26B + VRPRM 2025.08		53.55
InternVL2.5-8B + VRPRM 2025.08		51.52
InternVL2.5-38B + VRPRM w/o RL 2025.08		51.52
Gemini-2.0-Flash 2025.08		47.8
LUSPO 2026.02		47.1
GRPO 2026.02		46.8
InternVL2.5-38B + VisualPRM 2025.08		46.7
Claude-3.5-Sonnet 2025.08		46.3
GSPO 2026.02		45.6
InternVL2.5-26B + VRPRM w/o RL 2025.08		44.29
Qwen2.5-VL-7B-Instruct (w/o RLVR) 2026.02		41.1
GPT-4o 2025.08		40.6
InternVL2.5-8B + VRPRM w/o RL 2025.08		39.85
InternVL2.5-26B + VisualPRM 2025.08		39.1
InternVL2.5-38B 2025.08		36.9
InternVL2.5-8B + VisualPRM 2025.08		35.8
InternVL2.5-26B 2025.08		24
InternVL2.5-8B 2025.08		22.8