Region Captioning on ParaDLC-Bench
49Positive RateGAR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GARSize=8B, TPF=1, Time (s)=4792026.06 | 49 | 87.6 | 69.5 | |
| DAMSize=3B, TPF=1, Time (s)=3262026.06 | 48.1 | 87.2 | 69.2 | |
| Gemini-3.1-ProSize=-, TPF=-, Time (s)=-2026.06 | 43.6 | 81.1 | 63.7 | |
| PerceptionDLMSize=8B, TPF=2.9, Time (s)=2762026.06 | 42.3 | 82.4 | 62.4 | |
| PixelReferSize=7B, TPF=1, Time (s)=7182026.06 | 40.8 | 78.7 | 60.5 | |
| Gemini-2.5-ProSize=-, TPF=-, Time (s)=-2026.06 | 39.7 | 73.3 | 57.5 | |
| GPT-5.2Size=-, TPF=-, Time (s)=-2026.06 | 38 | 71 | 55.2 | |
| SDAR-VLSize=8B, TPF=1*, Time (s)=9452026.06 | 30.2 | 28.8 | 31.3 | |
| Dream-VLSize=7B, TPF=1*, Time (s)=4462026.06 | 29.7 | 28.6 | 30.4 | |
| LLaDA-VSize=8B, TPF=1*, Time (s)=32412026.06 | 24.1 | 46.3 | 35.2 |