Out-of-Distribution General Visual Question Answering on LLaVA-W
0.755ScoreMM1.5 1B (MoE)
Evaluation results
| Method | Links | |
|---|---|---|
| MM1.5 1B (MoE)Model Parameters=1B, Architecture=MoE, Decoding Strategy=Greedy2024.09 | 0.755 | |
| MM1.5 3BModel Parameters=3B, Decoding Strategy=Greedy2024.09 | 0.73 | |
| MM1.5 1BModel Parameters=1B, Decoding Strategy=Greedy2024.09 | 0.716 | |
| Phi-3-Vision 4BModel Parameters=4B, Decoding Strategy=Greedy2024.09 | 0.716 | |
| MiniCPM-V2 3BModel Parameters=3B, Decoding Strategy=Beam Search2024.09 | 0.692 | |
| InternVL2 2BModel Parameters=2B, Decoding Strategy=Beam Search2024.09 | 0.6 |