Multimodal Reasoning on ChartMuseum (Accuracy)
31.3AccuracyDUPL
Evaluation results
| Method | Links | |
|---|---|---|
| DUPLBackbone=Qwen2.5-VL-7B, Training Algorithm=DUPL2025.10 | 31.3 | |
| GRPOBackbone=Qwen2.5-VL-7B, Training Algorithm=GRPO2025.10 | 28.8 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2025.10 | 26.8 |