Object Counting Ranking on COCO-REM 8-imgs
0.823Kendall's TauQwen2.5-VL-7B + RARL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL-7B + RARLBackbone=Qwen2.5-VL-7B, Training Strategy=RARL2026.01 | 0.823 | |
| Qwen2.5-VL-3B + RARLBackbone=Qwen2.5-VL-3B, Training Strategy=RARL2026.01 | 0.797 | |
| Qwen2.5-VL-3B + SFTBackbone=Qwen2.5-VL-3B, Training Strategy=SFT2026.01 | 0.764 | |
| Qwen2.5-VL-7B + SFTBackbone=Qwen2.5-VL-7B, Training Strategy=SFT2026.01 | 0.763 | |
| Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B, Training Strategy=Base2026.01 | 0.74 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B, Training Strategy=Base2026.01 | 0.721 |