Visual Hallucination Evaluation on HallusionBench visual questions
65.8AccuracyGPT-4V
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4VType=Closed-source API2024.01 | 65.8 | |
| Gemini-ProType=Closed-source API2024.01 | 63.9 | |
| InternLM-XComposer2LLM=InernLM2-7B2024.01 | 60.3 | |
| InternLM-XCLLM=InernLM-7B2024.01 | 57 | |
| QwenVL-PlusType=Closed-source API2024.01 | 56.4 | |
| Qwen-VL-ChatLLM=Qwen-7B2024.01 | 56.4 | |
| InstructBLIPLLM=Vicuna-7B2024.01 | 53.6 | |
| LLaVA-1.5LLM=Vicuna-13B2024.01 | 46.7 | |
| IDEFICS-80BLLM=LLaMA-65B2024.01 | 46.1 | |
| LLaVALLM=Vicuna-7B2024.01 | 44.1 |