Image Description Hallucination Evaluation on Visual Genome (test)
37.2SHRLLaVA-1.5-13B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-1.5-13BMax new tokens=642025.04 | 37.2 | |
| LLaVA-1.5-7BMax new tokens=642025.04 | 36.7 | |
| SeVaBackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 34.9 | |
| VCDBackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 34.2 | |
| OPERABackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 34.1 | |
| HA-DPOBackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 34 | |
| SIDBackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 33.1 | |
| LLaVA-Next-Mistral-7BMax new tokens=642025.04 | 28.4 | |
| AntidoteBackbone=LLaVA-1.5-13B, Max new tokens=642025.04 | 21.3 | |
| AntidoteBackbone=LLaVA-Next-Mistral-7B, Max new tokens=642025.04 | 19.7 | |
| AntidoteBackbone=LLaVA-1.5-7B, Max new tokens=642025.04 | 18.1 |