Fine-grained Visual Perception on MagnifierBench
42.7Multi-ScoreOtterHD-8B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OtterHD-8BI/R Pairs=370K, Train Res.=Dynamic, Eval Res.=Original2023.11 | 42.7 | 39.9 | |
| OtterHD-8BI/R Pairs=370K, Train Res.=1024, Eval Res.=10242023.11 | 37.8 | 37.2 | |
| OtterHD-8BI/R Pairs=370K, Train Res.=512, Eval Res.=5122023.11 | 33.5 | 31.4 | |
| Fuyu-8BEval Res.=Original2023.11 | 29.3 | 15.2 | |
| LLaVA-7B1.5I/R Pairs=3.6M, Train Res.=336, Eval Res.=3362023.11 | 26.8 | 24.7 | |
| Otter-9BI/R Pairs=150K, Train Res.=224, Eval Res.=2242023.11 | 25.7 | 15.8 | |
| Idefics-9BinstructI/R Pairs=1M, Train Res.=224, Eval Res.=2242023.11 | 20.8 | 13.4 | |
| Qwen-VL-7BchatI/R Pairs=1.4B, Train Res.=448, Eval Res.=4482023.11 | 14.5 | 15.9 | |
| InstructBLIP-7BI/R Pairs=1.2M, Train Res.=224, Eval Res.=2242023.11 | 5.6 | 15.2 | |
| InstructBLIP-13BI/R Pairs=1.2M, Train Res.=224, Eval Res.=2242023.11 | 3.8 | 16.3 |