Multi-view Hallucination Evaluation on MVH-Bench Cross-View 1.0 (test)
72.87Binary AccuracyRSCD
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RSCDBackbone=Qwen2.5-VL-7B2026.03 | 72.87 | 56.02 | 31.02 | 77.36 | 58.33 | 295.6 | |
| ICDBackbone=Qwen2.5-VL-7B2026.03 | 71.18 | 53.34 | 27.97 | 76.06 | 56.48 | 285.03 | |
| VCDBackbone=Qwen2.5-VL-7B2026.03 | 70.37 | 52.92 | 27.41 | 75.05 | 54.81 | 280.55 | |
| AvisCBackbone=Qwen2.5-VL-7B2026.03 | 69.28 | 49.68 | 24.35 | 74.77 | 54.82 | 272.9 | |
| Base modelBackbone=Qwen2.5-VL-7B2026.03 | 67.45 | 47.45 | 21.95 | 70.7 | 50.56 | 258.11 | |
| RSCDBackbone=LLaVA-OneVision-7B2026.03 | 56.57 | 34.31 | 5.19 | 63.15 | 30.83 | 190.05 | |
| VCDBackbone=LLaVA-OneVision-7B2026.03 | 54.91 | 29.95 | 2.59 | 62.31 | 30.28 | 180.04 | |
| ICDBackbone=LLaVA-OneVision-7B2026.03 | 53.98 | 26.95 | 2.5 | 60.51 | 28.61 | 172.55 | |
| AvisCBackbone=LLaVA-OneVision-7B2026.03 | 53.64 | 26.39 | 1.57 | 61.71 | 28.42 | 171.74 | |
| Base modelBackbone=LLaVA-OneVision-7B2026.03 | 52.36 | 27.17 | 4.26 | 59.26 | 29.72 | 172.77 |