Visual Question Answering on GQA balanced (test)
66.6AccuracyInternVL-Chat
Evaluation results
| Method | Links | |
|---|---|---|
| InternVL-ChatLLM=Vicuna-13B, glue layer=QLLAMA2023.12 | 66.6 | |
| InternVL-ChatLLM=Vicuna-13B, glue layer=MLP2023.12 | 63.9 | |
| LLaVA-1.5LLM=Vicuna-13B2023.12 | 63.3 | |
| InternVL-ChatLLM=Vicuna-7B, glue layer=MLP2023.12 | 62.9 | |
| LLaVA-1.5LLM=Vicuna-7B2023.12 | 62 | |
| InternVL-ChatLLM=Vicuna-13B, Res.=224, glue layer=QLLAMA2023.12 | 59.5 | |
| Qwen-VLLLM=Qwen-7B2023.12 | 59.3 | |
| InternVL-ChatLLM=Vicuna-7B, Res.=224, glue layer=QLLAMA2023.12 | 57.7 | |
| Qwen-VL-ChatLLM=Qwen-7B2023.12 | 57.5 | |
| IDEFICS-80BLLM=LLaMA-65B2023.12 | 45.2 | |
| BLIP-2LLM=Vicuna-13B2023.12 | 41 | |
| InstructBLIPLLM=Vicuna-7B2023.12 | 34.5 |