Pairwise Ranking on WildVision Arena in-domain
62Accuracy (w/ Tie)GPT-4V
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4V2024.10 | 62 | 73.3 | 0.787 | |
| GPT-4o2024.10 | 61.7 | 73.4 | 0.819 | |
| LLaVA-Critic-72BParameters=72B2024.10 | 60.5 | 73.6 | 0.779 | |
| LLaVA-Critic-7BParameters=7B2024.10 | 59.6 | 72.2 | 0.763 | |
| LLaVA-OV-72BParameters=72B2024.10 | 59.4 | 70.8 | 0.763 | |
| LLaVA-Critic (LLaVA-NeXT)Backbone=LLaVA-NeXT2024.10 | 58.2 | 68.6 | 0.755 | |
| LLaVA-Critic-7B (v0.5)Parameters=7B, Version=v0.52024.10 | 58 | 69.2 | 0.755 | |
| Qwen2-VL-7B-InstructParameters=7B, Type=Instruct2024.10 | 55 | 67.8 | 0.699 | |
| LLaVA-OV-7BParameters=7B2024.10 | 53.1 | 64 | 0.715 | |
| LLaMA3.2-VParameters=11B, Type=Instruct2024.10 | 51.3 | 67.3 | 0.737 | |
| LLaVA-NeXTBackbone=LLaMA-8B2024.10 | 47.3 | 56.9 | 0.605 |