Referring Text Classification on COCO-Text Out-of-Domain
84.62AccuracyLLaVA-1.5 + Blur
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-1.5 + BlurInput format=Box, Training strategy=Training-Free Methods2026.02 | 84.62 | |
| LLaVA-1.5 + BlurInput format=Mask, Training strategy=Training-Free Methods2026.02 | 75.85 | |
| LLaVA-1.5 + ControlMLLM++Input format=Box, Training strategy=Training-Free Methods2026.02 | 74.66 | |
| LLaVA-1.5 + ControlMLLM++Input format=Mask, Training strategy=Training-Free Methods2026.02 | 74.65 | |
| LLaVA-1.5 + ControlMLLMInput format=Box, Training strategy=Training-Free Methods2026.02 | 63.06 | |
| LLaVA-1.5 + ControlMLLMInput format=Mask, Training strategy=Training-Free Methods2026.02 | 61.43 | |
| Shikra-7BInput format=Box, Training strategy=Training Methods2026.02 | 60.41 | |
| Ferret-7BInput format=Mask, Training strategy=Training Methods2026.02 | 59.81 | |
| LLaVA-1.5 + ColorInput format=Box, Training strategy=Training-Free Methods2026.02 | 59.04 | |
| Ferret-7BInput format=Box, Training strategy=Training Methods2026.02 | 58.7 | |
| LLaVA-1.5Input format=Box, Training strategy=Training-Free Methods2026.02 | 57.42 | |
| LLaVA-1.5Input format=Mask, Training strategy=Training-Free Methods2026.02 | 57.42 | |
| GPT4-ROIInput format=Box, Training strategy=Training Methods2026.02 | 57.17 | |
| LLaVA-1.5 + ColorInput format=Mask, Training strategy=Training-Free Methods2026.02 | 56.57 | |
| LLaVA-1.5 + Edit AttInput format=Mask, Training strategy=Training-Free Methods2026.02 | 51.88 | |
| LLaVA-1.5 + Edit AttInput format=Box, Training strategy=Training-Free Methods2026.02 | 50 | |
| Kosmos-2Input format=Box, Training strategy=Training Methods2026.02 | 23.12 |