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

Multimodal Mathematical Reasoning on MathVerse-V

81.2Accuracy

Gemini-2.5-Pro-Thinking

Updated 2mo ago

Evaluation Results

Method	Links
Gemini-2.5-Pro-Thinking 2025.12		81.2
Seed-1.5-thinking 2025.12		80.4
Qwen2.5-VL-7B + PGPO 2026.04		66.41
Claude-4-Sonnet 2025.12		66.1
Qwen2.5-VL-7B + VPPO 2026.04		65.54
VL-Rethinker-7B 2026.04		64.96
Qwen2.5-VL-7B + PAPO 2026.04		64.92
Qwen2.5-VL-7B + GRPO 2026.04		64.78
Qwen2.5-VL-7B + DAPO 2026.04		64.38
R1-ShareVL-7B 2026.04		64.34
NoisyRollout-7B 2026.04		63.78
MM-Eureka-7B 2026.04		62.41
GPT-4.1 2025.12		59.8
Qwen2.5-VL-3B + PGPO 2026.04		57.24
Qwen2.5-VL-3B + VPPO 2026.04		57.02
Qwen2.5-VL-3B + PAPO 2026.04		56.9
Qwen2.5-VL-3B + DAPO 2026.04		54.89
Qwen2.5-VL-3B + GRPO 2026.04		54.36
Revisual-R1 2025.12		53.6
VAPO-Thinker 2025.12		53.3
Ours 2025.12		53.3
Look-Back 2025.12		51.8
Vision-G1 2025.12		50
MM-Eureka 2025.12		49.6
VL-Rethinker 2025.12		49.1
VLAA-Thinking 2025.12		48.2
ThinkLite-VL 2025.12		45.3
MathBook 2025.12		45.2
MMR1-Math-v0 2025.12		45.1
Qwen2.5-VL 2025.12		42.9
OpenVLThinker 2025.12		40.7
Qwen2.5-VL-7B 2026.04		33.63
Qwen2.5-VL-3B 2026.04		30.31