Open-ended Visual Question Answering on LLaVA Eval v1 (test)
77.67Conversation ScoreDRESS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DRESSprefix=<excellent> [Nice response.]2023.11 | 77.67 | 62.17 | 84.27 | 74.7 | |
| InstructBLIPLLM=Vicuna-13B2023.11 | 74.08 | 61.67 | 82.17 | 72.64 | |
| LLaVA-HFLLM=Vicuna-13B2023.11 | 69.74 | 60.87 | 85.33 | 71.98 | |
| BLIP-2LLM=T5-XXL2023.11 | 66.08 | 31.33 | 22 | 39.8 | |
| mPLUGLLM=LLaMA-7B2023.11 | 66.08 | 44.17 | 75.83 | 62.03 | |
| LLaVALLM=LLaMA-13B2023.11 | 65.17 | 42.17 | 61.5 | 56.28 | |
| miniGPT4LLM=Vicuna-13B2023.11 | 54.92 | 51.5 | 74.67 | 60.36 |