Visual perception and grounding on HRBench
74.2AccuracyMoCA
Evaluation results
| Method | Links | |
|---|---|---|
| MoCASize=7B2026.05 | 74.2 | |
| Qwen2.5-VL-InstructSize=72B2026.05 | 73.4 | |
| DeepEyesSize=7B2026.05 | 73.1 | |
| Pixel ReasonerSize=7B2026.05 | 72.8 | |
| Qwen2.5-VL-InstructSize=7B2026.05 | 69.2 | |
| GPT-4oSize=-2026.05 | 65 | |
| VL-RethinkerSize=7B2026.05 | 65 | |
| Llava-OVSize=7B2026.05 | 64.7 | |
| mPLUG-Owl3Size=7B2026.05 | 60.1 | |
| R1-VLSize=7B2026.05 | 53.1 | |
| DocopilotSize=8B2026.05 | 48.3 | |
| GPT-4o-miniSize=-2026.05 | 48 |