Image-based Reasoning on MM-IT 1K (test)
58Response AccuracyAnyMAL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AnyMALLLM Size=70B, Training Data=MM-IT Human+Synth2023.09 | 58 | 79.3 | 99.7 | |
| AnyMALLLM Size=70B2023.09 | 56 | 82.4 | 99.3 | |
| AnyMALLLM Size=70B, Training Data=MM-IT Synth Only2023.09 | 54.2 | 83.5 | 99.5 | |
| LLaVA2023.09 | 51.7 | 85.4 | 99.5 | |
| InstructBLIPBackbone=Vicuna-13B2023.09 | 46.3 | 73.2 | 98.3 | |
| BLIP-2Backbone=FlanT5_XXL2023.09 | 43.3 | 73.5 | 99.3 | |
| Mini-GPT42023.09 | 42.7 | 73 | 99.5 |