Visual Reasoning on ScienceQA
87.2AccuracyDualMindVLM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DualMindVLMSize=7B, Strategy=RL2025.11 | 87.2 | 98 | — | |
| Supervised ETTCEnsemble Composition=Same Family Models2026.05 | 86.07 | — | 1.04 | |
| Supervised ETTCEnsemble Composition=Similar Size Models2026.05 | 85.97 | — | 0.69 | |
| VotingEnsemble Composition=Similar Size Models2026.05 | 85.52 | — | — | |
| VL-RethinkerSize=7B, Strategy=RL2025.11 | 85.5 | 205 | — | |
| ETTCEnsemble Composition=Similar Size Models2026.05 | 85.28 | — | — | |
| ETTCEnsemble Composition=Same Family Models2026.05 | 85.03 | — | — | |
| VotingEnsemble Composition=Same Family Models2026.05 | 84.04 | — | — | |
| Qwen2.5-VLSize=7B, Strategy=-2025.11 | 84 | 156 | — | |
| MM-EurekaSize=7B, Strategy=RL2025.11 | 83.5 | 202 | — | |
| OpenVLThinkerSize=7B, Strategy=SFT+RL2025.11 | 82.2 | 171 | — | |
| Average (Single-Model)Ensemble Composition=Same Family Models2026.05 | 78.82 | — | — | |
| Average (Single-Model)Ensemble Composition=Similar Size Models2026.05 | 78.44 | — | — |