Hallucination Robustness on HallusionBench (fAcc)
37.3fAccQwen2.5-VL + DRScaffold
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL + DRScaffoldScale=3B2026.05 | 37.3 | |
| Qwen2.5-VLScale=3B2026.05 | 35.2 | |
| Phi4-multimodal + DRScaffoldScale=5.6B2026.05 | 34.4 | |
| InternVL2.5 + DRScaffoldScale=2B2026.05 | 33.8 | |
| InternVL2.5Scale=2B2026.05 | 32.6 | |
| Qwen2.5-VLScale=32B2026.05 | 31.2 | |
| Phi4-multimodalScale=5.6B2026.05 | 25.1 |