High-resolution Visual Question Answering on HR-Bench 8K
76.9Overall AccuracyPFlowNet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PFlowNet2026.05 | 76.9 | 89 | 64.8 | |
| Qwen2.5-VL-72B2026.05 | 76.3 | 84.3 | 68.3 | |
| Qwen3-VL-32B2026.05 | 74.8 | 90.1 | 59.5 | |
| DeepEyesV22026.05 | 73.8 | 88.5 | 59 | |
| TreeVGR2026.05 | 72.8 | 86 | 59.5 | |
| DeepEyes2026.05 | 72.6 | 86.8 | 58.5 | |
| Thyme2026.05 | 72 | 86.5 | 57.5 | |
| Qwen2.5-VL-32B2026.05 | 71.6 | 86.5 | 56.8 | |
| Qwen2.5-VL-7B2026.05 | 68.8 | 83.5 | 54 | |
| Qwen3-VL-8B2026.05 | 68.1 | 82 | 54.3 | |
| Qwen3-VL-4B2026.05 | 67.1 | 83.5 | 50.7 | |
| InternVL3-38B2026.05 | 67 | 71.3 | 62.8 | |
| PixelReasoner2026.05 | 66.9 | 80 | 54.3 | |
| InternVL3-8B2026.05 | 62 | 64.3 | 59.8 | |
| LLaVA-OV-72B2026.05 | 60.9 | 68.8 | 53 | |
| GPT-4o-11202026.05 | 55.5 | 62 | 49 |