Multi-image Understanding and Reasoning on LLaVA-Interleave Bench In-domain
62.3Avg ScoreLLaVA-NeXT-Interleave
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LLaVA-NeXT-InterleaveParameters=14B2024.07 | 62.3 | 40.5 | 24.5 | 33.3 | 78.6 | 95 | 59.9 | 76.7 | 91.1 | |
| LLaVA-NeXT-InterleaveParameters=7B2024.07 | 58.6 | 37.1 | 24.3 | 33.1 | 76.1 | 87.5 | 48.7 | 74.2 | 88.8 | |
| LLaVA-NeXT-InterleaveParameters=0.5B2024.07 | 43.9 | 34.3 | 21.6 | 29.7 | 63.9 | 54.8 | 35.4 | 52 | 67.8 | |
| MantisParameters=7B2024.07 | 39.6 | 17.6 | 11.2 | 12.5 | 45.2 | 52.5 | 25.7 | 69.9 | 87.4 | |
| GPT-4V2024.07 | 39.2 | 12.5 | 11 | 10.9 | 54.5 | 52 | 17.1 | 76.5 | 88.8 | |
| VPG-CParameters=7B2024.07 | 35.8 | 27.8 | 15.2 | 21.5 | 38.9 | 46.8 | 2.4 | 57.6 | 73.2 | |
| LLaVA-NeXT-ImageParameters=7B2024.07 | 32.4 | 12.9 | 13.2 | 10.1 | 59.6 | 39.4 | 9 | 51 | 68 |