Multimodal Understanding on MMBench E 80% (dev test)
86.67AccuracyQwen2-VL-SCI5
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2-VL-SCI5Base Model=Qwen2-VL, Inference Strategy=SCI52026.03 | 86.67 | |
| Qwen2-VL-TIEBase Model=Qwen2-VL, Inference Strategy=TIE2026.03 | 86.59 | |
| Qwen2-VL-VCDBase Model=Qwen2-VL, Inference Strategy=VCD2026.03 | 86.56 | |
| Qwen2-VL-M3IDBase Model=Qwen2-VL, Inference Strategy=M3ID2026.03 | 86.46 | |
| Qwen2-VLBase Model=Qwen2-VL2026.03 | 86.36 | |
| LLaVA-NeXT-TIEBase Model=LLaVA-NeXT, Inference Strategy=TIE2026.03 | 80.28 | |
| LLaVA-NeXT-VCDBase Model=LLaVA-NeXT, Inference Strategy=VCD2026.03 | 80.28 | |
| LLaVA-NeXT-M3IDBase Model=LLaVA-NeXT, Inference Strategy=M3ID2026.03 | 80.18 | |
| LLaVA-NeXT-SCI5Base Model=LLaVA-NeXT, Inference Strategy=SCI52026.03 | 80.08 | |
| LLaVA-NeXTBase Model=LLaVA-NeXT2026.03 | 79.72 |