Reasoning over rich modalities on MMLongBench Doc
42.3AccuracyGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oSize=-2026.05 | 42.3 | |
| MoCASize=7B2026.05 | 33.1 | |
| DocopilotSize=8B2026.05 | 28.8 | |
| GPT-4o-miniSize=-2026.05 | 28.6 | |
| Qwen2.5-VL-InstructSize=72B2026.05 | 24.9 | |
| Pixel ReasonerSize=7B2026.05 | 22 | |
| Qwen2.5-VL-InstructSize=7B2026.05 | 21.2 | |
| mPLUG-Owl3Size=7B2026.05 | 21 | |
| VL-RethinkerSize=7B2026.05 | 20.9 | |
| Llava-OVSize=7B2026.05 | 19.5 | |
| DeepEyesSize=7B2026.05 | 17.5 | |
| R1-VLSize=7B2026.05 | 12.7 |