Multimodal Visual Question Answering on MathVision
27.18AccuracySPINE
Evaluation results
| Method | Links | |
|---|---|---|
| SPINEBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 27.18 | |
| TTRLBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 22.73 | |
| No adaptationBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 19.65 | |
| Self-ConsistencyBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 19.2 | |
| SEALONGBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 10.36 | |
| LMSIBase Model=Qwen2.5-VL-3B-Instruct2025.11 | 9.21 |