General Multi-image Reasoning and Generalization on LLaVA-Interleave Bench Out-domain
57.8Average ScoreGPT-4V
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-4V2024.07 | 57.8 | 60.3 | 66.9 | 62.7 | 51.1 | 47.9 | — | |
| LLaVA-NeXT-InterleaveParameters=14B2024.07 | 44.3 | 33.4 | 32.7 | 66.4 | 52.1 | 37.1 | — | |
| LLaVA-NeXT-InterleaveParameters=7B2024.07 | 42.8 | 32.8 | 31.6 | 62.7 | 52.6 | 34.5 | — | |
| MantisParameters=7B2024.07 | 39.3 | 27.2 | 29.3 | 59.5 | 46.4 | 34.1 | — | |
| VPG-CParameters=7B2024.07 | 34.5 | 24.3 | 23.1 | 52.4 | 43.1 | 29.4 | — | |
| LLaVA-NeXT-InterleaveParameters=0.5B2024.07 | 33.1 | 13.3 | 12.2 | 45.6 | 39.2 | 28.6 | — | |
| LLaVA-NeXT-ImageParameters=7B2024.07 | 29.4 | 13.5 | 12.2 | 46.1 | 41.8 | 33.5 | — | |
| GPT-4VVersion=V-Preview2024.08 | — | — | — | — | — | — | 60.3 | |
| LLaVA-OneVision-0.5BModel Scale=0.5B2024.08 | — | — | — | — | — | — | 33.3 | |
| LLaVA-OneVision-72BModel Scale=72B2024.08 | — | — | — | — | — | — | 79.9 | |
| LLaVA-OneVision-7BModel Scale=7B2024.08 | — | — | — | — | — | — | 64.2 |