Vision-Language Evaluation on CV-Bench
90.1AccuracyVLSI-2B
Evaluation Results
| Method | Links | |
|---|---|---|
| VLSI-2BSize=2B2024.12 | 90.1 | |
| VLSI-7BSize=7B2024.12 | 89.1 | |
| Cambrian-34BSize=34B2024.12 | 76.9 | |
| Phantom-7BSize=7B2024.12 | 74.9 | |
| PhantomParameters=7B2024.09 | 74.9 | |
| Qwen2-VL-72BSize=72B2024.12 | 74.3 | |
| Florence-VL 8B# Vis tok.=5762024.12 | 73.4 | |
| Cambrian-1Parameters=8B2024.09 | 72.2 | |
| Cambrian 8B# Vis tok.=5762024.12 | 72.2 | |
| Florence-VL 3B# Vis tok.=5762024.12 | 70.2 | |
| Phi 3.5 Vision2024.12 | 69.3 | |
| GPT-4V (0409)Model Version=04092024.12 | 69.1 | |
| GPT-4V2024.09 | 69.1 | |
| LLaVA next 8B# Vis tok.=28802024.12 | 63.8 | |
| Phantom-1.8BSize=1.8B2024.12 | 63.1 | |
| Mini-Gemini-HD 8B# Vis tok.=28802024.12 | 62.6 | |
| Vila 3B2024.12 | 55.2 |