Visual Reasoning on MathVista mini (test)
77.6AccuracyInsight-V++
Evaluation results
| Method | Links | |
|---|---|---|
| Insight-V++Size=7B, Base Model=Qwen2.5-VL, Self-Evolving=true2026.03 | 77.6 | |
| Qwen2.5-VL + Multi-Agent (RL)Size=7B, Multi-Agent (RL)=true2026.03 | 74.5 | |
| Qwen2.5-VLSize=7B2026.03 | 69.2 | |
| IXC-2.5Size=7B2026.03 | 63.7 | |
| POINTSSize=7B2026.03 | 63 | |
| Ovis1.5-LLaMA3Size=8B2026.03 | 63 | |
| Qwen2-VLSize=7B2026.03 | 61.4 | |
| MiniCPM-V-2.6Size=7B2026.03 | 60.6 | |
| Insight-VSize=7B, Base Model=Our Base Model, Iterative DPO=true2026.03 | 59.9 | |
| InternLM-XComposer2Size=7B2026.03 | 59.5 | |
| Our Base Model + Multi-AgentSize=7B, Multi-Agent=true2026.03 | 58.7 | |
| Idefics3-LLaMA3Size=8B2026.03 | 58.4 | |
| Our Base ModelSize=7B2026.03 | 56.9 | |
| MiniCPM-LLaMA3-V 2.5Size=8B2026.03 | 54.3 | |
| Insight-V-LLaVASize=8B, Base Model=LLaVA-NeXT-LLaMA3, Iterative DPO=true2026.03 | 49.8 | |
| Cambrian-1Size=8B2026.03 | 49 | |
| MM-1.5Size=7B2026.03 | 47.6 | |
| LLaVA-NeXT-LLaMA3 + Multi-AgentSize=8B, Multi-Agent=true2026.03 | 47.4 | |
| LLaVA-NeXT-LLaMA3Size=8B2026.03 | 45.9 | |
| DeepSeek-VLSize=7B2026.03 | 36.1 | |
| Bunny-LLaMA3Size=8B2026.03 | 34.4 |