Hallucination / grounding Accuracy (aAcc) on HallusionBench
72.9Answer Accuracy (aAcc)S^3-FT
Evaluation Results
| Method | Links | |
|---|---|---|
| S^3-FTBackbone=Qwen3-VL-8B, Mode=Single-view2026.05 | 72.9 | |
| S^3-FTBackbone=Qwen3-VL-8B, Mode=Cross-viewpoint2026.05 | 72.5 | |
| S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Cross-viewpoint2026.05 | 69.4 | |
| S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Single-view2026.05 | 68.8 | |
| Qwen2.5-VL-7BMode=base2026.05 | 65.2 | |
| Qwen3-VL-8BMode=base2026.05 | 55.8 |