Visual Hallucination Evaluation on POPE (average)
88.93F1 ScoreCRG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CRGBackbone=Qwen2.5-VL-32B-Instruct2026.05 | 88.93 | 89.92 | |
| CRGBackbone=Qwen2.5-VL-7B-Instruct2026.05 | 88.55 | 89.72 | |
| CRGBackbone=LLaVA-1.5-13B2026.05 | 87.94 | 87.87 | |
| CRGBackbone=LLaVA-NeXT-34B2026.05 | 87.73 | 89.74 | |
| InternLM-XComposer2LLM=InernLM2-7B2024.01 | 87.7 | — | |
| CRGBackbone=LLaVA-1.5-7B2026.05 | 86.61 | 87.71 | |
| LLaVA-1.5LLM=Vicuna-13B2024.01 | 85.9 | — | |
| RegularBackbone=Qwen2.5-VL-32B-Instruct2026.05 | 85.31 | 84.25 | |
| RegularBackbone=LLaVA-NeXT-34B2026.05 | 84.79 | 83.17 | |
| RegularBackbone=Qwen2.5-VL-7B-Instruct2026.05 | 84.31 | 83.99 | |
| RegularBackbone=LLaVA-1.5-13B2026.05 | 82.81 | 83.05 | |
| LLaVALLM=Vicuna-7B2024.01 | 80.2 | — | |
| RegularBackbone=LLaVA-1.5-7B2026.05 | 79.65 | 81.37 | |
| InstructBLIPLLM=Vicuna-7B2024.01 | 78.9 | — |