Open-ended Visual Question Answering on LLaVA Bench v1 (test)
37.18RelevanceDRESS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DRESSprefix=<excellent> [Nice response.]2023.11 | 37.18 | 20.12 | 21.87 | 26.45 | |
| mPLUGLLM=LLaMA-7B2023.11 | 35.17 | 20.33 | 16.33 | 20.33 | |
| LLaVA-HFLLM=Vicuna-13B2023.11 | 34.33 | 18.5 | 17.67 | 23.5 | |
| InstructBLIPLLM=Vicuna-13B2023.11 | 34 | 21 | 19.67 | 22.67 | |
| miniGPT4LLM=Vicuna-13B2023.11 | 32.45 | 20.33 | 20.17 | 24.17 | |
| LLaVALLM=LLaMA-13B2023.11 | 31.83 | 19.83 | 18.67 | 20.83 | |
| BLIP-2LLM=T5-XXL2023.11 | 25 | 16 | 16 | 17.67 |