Open-ended Visual Chat on LLaVA-Bench In-the-Wild (full)
90.1Reasoning ScoreBing-Chat-0629
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Bing-Chat-06292023.09 | 90.1 | 59.6 | 52.2 | 71.5 | |
| LLaVA-65Bbeam search size=52023.09 | 88.7 | 59.4 | 65.7 | 74.4 | |
| LLaVA-65Bbeam search size=12023.09 | 87.3 | 63.8 | 62.3 | 74.2 | |
| LLaVA-13Bbeam search size=52023.09 | 84.3 | 68.4 | 59.9 | 73.5 | |
| LLaVA-33Bbeam search size=52023.09 | 83.5 | 72.6 | 61.9 | 74.8 | |
| LLaVA-33Bbeam search size=12023.09 | 82.9 | 70.2 | 62.6 | 73.9 | |
| LLaVA-13Bbeam search size=12023.09 | 81.7 | 64.3 | 55.9 | 70.1 | |
| Bard-07182023.09 | 78.7 | 83.7 | 69.7 | 77.8 |