Multimodal Mathematical Reasoning on MathVerse-V
81.2AccuracyGemini-2.5-Pro-Thinking
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-2.5-Pro-ThinkingAccess Type=Closed-Source2025.12 | 81.2 | |
| Seed-1.5-thinkingAccess Type=Closed-Source2025.12 | 80.4 | |
| Qwen2.5-VL-7B + PGPOModel Size=7B, Optimization Strategy=PGPO2026.04 | 66.41 | |
| Claude-4-SonnetAccess Type=Closed-Source2025.12 | 66.1 | |
| Qwen2.5-VL-7B + VPPOModel Size=7B, Optimization Strategy=VPPO2026.04 | 65.54 | |
| VL-Rethinker-7BModel Size=7B2026.04 | 64.96 | |
| Qwen2.5-VL-7B + PAPOModel Size=7B, Optimization Strategy=PAPO2026.04 | 64.92 | |
| Qwen2.5-VL-7B + GRPOModel Size=7B, Optimization Strategy=GRPO2026.04 | 64.78 | |
| Qwen2.5-VL-7B + DAPOModel Size=7B, Optimization Strategy=DAPO2026.04 | 64.38 | |
| R1-ShareVL-7BModel Size=7B2026.04 | 64.34 | |
| NoisyRollout-7BModel Size=7B2026.04 | 63.78 | |
| MM-Eureka-7BModel Size=7B2026.04 | 62.41 | |
| GPT-4.1Access Type=Closed-Source2025.12 | 59.8 | |
| Qwen2.5-VL-3B + PGPOModel Size=3B, Optimization Strategy=PGPO2026.04 | 57.24 | |
| Qwen2.5-VL-3B + VPPOModel Size=3B, Optimization Strategy=VPPO2026.04 | 57.02 | |
| Qwen2.5-VL-3B + PAPOModel Size=3B, Optimization Strategy=PAPO2026.04 | 56.9 | |
| Qwen2.5-VL-3B + DAPOModel Size=3B, Optimization Strategy=DAPO2026.04 | 54.89 | |
| Qwen2.5-VL-3B + GRPOModel Size=3B, Optimization Strategy=GRPO2026.04 | 54.36 | |
| Revisual-R1Access Type=Open-Source, Model Scale=7B2025.12 | 53.6 | |
| VAPO-ThinkerAccess Type=Open-Source, Model Scale=7B2025.12 | 53.3 | |
| OursAccess Type=Open-Source, Model Scale=7B2025.12 | 53.3 | |
| Look-BackAccess Type=Open-Source, Model Scale=7B2025.12 | 51.8 | |
| Vision-G1Access Type=Open-Source, Model Scale=7B2025.12 | 50 | |
| MM-EurekaAccess Type=Open-Source, Model Scale=7B2025.12 | 49.6 | |
| VL-RethinkerAccess Type=Open-Source, Model Scale=7B2025.12 | 49.1 | |
| VLAA-ThinkingAccess Type=Open-Source, Model Scale=7B2025.12 | 48.2 | |
| ThinkLite-VLAccess Type=Open-Source, Model Scale=7B2025.12 | 45.3 | |
| MathBookAccess Type=Open-Source, Model Scale=7B2025.12 | 45.2 | |
| MMR1-Math-v0Access Type=Open-Source, Model Scale=7B2025.12 | 45.1 | |
| Qwen2.5-VLAccess Type=Open-Source, Model Scale=7B2025.12 | 42.9 | |
| OpenVLThinkerAccess Type=Open-Source, Model Scale=7B2025.12 | 40.7 | |
| Qwen2.5-VL-7BModel Size=7B2026.04 | 33.63 | |
| Qwen2.5-VL-3BModel Size=3B2026.04 | 30.31 |