Multimodal Comprehension on Hallusion
50.9ScoreQwen2-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2-VLModel Size=7B, Training Strategy=Original2026.02 | 50.9 | |
| CREM_GModel Size=7B, Training Strategy=Fine-tuned on ShareGPT-4V2026.02 | 49.2 | |
| CREMModel Size=7B, Training Strategy=Unified framework2026.02 | 48.8 | |
| CREM_RModel Size=7B, Training Strategy=Trained on MMEB2026.02 | 44.5 | |
| Qwen2-VLModel Size=2B, Training Strategy=Original2026.02 | 41.9 | |
| CREMModel Size=2B, Training Strategy=Unified framework2026.02 | 41.2 | |
| CREM_GModel Size=2B, Training Strategy=Fine-tuned on ShareGPT-4V2026.02 | 41 | |
| CREM_RModel Size=2B, Training Strategy=Trained on MMEB2026.02 | 33.6 |