Object Hallucination Evaluation on MSCOCO (Unspecified Split)
56.4CHAIR Scene ScoreEAH
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EAHBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 56.4 | 9.9 | |
| HALCBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 49.4 | 14.2 | |
| EAHBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 47.5 | 13.2 | |
| OPERABackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 46.7 | 14.6 | |
| OPERABackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 45.2 | 12.4 | |
| GIFTBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 38.8 | 11.9 | |
| HALCBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 38.3 | 15.9 | |
| EAHBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 37.8 | 10.4 | |
| GIFTBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 37.4 | 9.6 | |
| OPERABackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 37.2 | 12.3 | |
| HALCBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 36.9 | 9.8 | |
| LVLMs-SaliencyBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 36.2 | 8.8 | |
| LVLMs-SaliencyBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 35.3 | 10.2 | |
| GIFTBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 33.5 | 11.7 | |
| DML-LVLMBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 32.5 | 13.8 | |
| LVLMs-SaliencyBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 31.6 | 10.2 | |
| EAHBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 30.7 | 9.9 | |
| GIFTBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 29.3 | 10.2 | |
| HALCBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 28.5 | 10.3 | |
| LVLMs-SaliencyBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 27.4 | 9.1 | |
| OPERABackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 26 | 9.1 | |
| DML-LVLMBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 25.7 | 7.2 | |
| SAGEBackbone=InstructBLIP, Maximum Generation Length=Averaged 256 and 5122026.03 | 25.7 | 9.4 | |
| DML-LVLMBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 25.7 | 7.2 | |
| DML-LVLMBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 24.6 | 9.9 | |
| SAGEBackbone=Shikra, Maximum Generation Length=Averaged 256 and 5122026.03 | 24.3 | 8.7 | |
| SAGEBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 23.8 | 8.2 | |
| OPERABackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 23.7 | 7.1 | |
| DML-LVLMBackbone=MiniGPT-4, Maximum Generation Length=Averaged 256 and 5122026.03 | 23.5 | 8.8 | |
| HALCBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 22.5 | 6.9 | |
| GIFTBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 22.5 | 8.1 | |
| EAHBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 22.4 | 6.5 | |
| SAGEBackbone=LLaVA-1.5, Maximum Generation Length=Averaged 256 and 5122026.03 | 21.3 | 5.2 | |
| SAGEBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 21.1 | 5.9 | |
| LVLMs-SaliencyBackbone=Qwen2-VL-7B, Maximum Generation Length=Averaged 256 and 5122026.03 | 20.1 | 5.4 |