Multimodal Question Answering on V* Bench
80.6Answer AccuracyHART-7B
Evaluation results
| Method | Links | |
|---|---|---|
| HART-7BResolution=2, 246 × 1, 5832026.02 | 80.6 | |
| Qwen2.5-VL-7BResolution=2, 246 × 1, 5832026.02 | 78.5 | |
| InternVL3-7BResolution=2, 246 × 1, 5832026.02 | 72.3 | |
| LLaVA-OneVision-7BResolution=2, 246 × 1, 5832026.02 | 70.7 |