Hallucination Evaluation on HaloQuest
80.8Score SQwen2.5-VL + FINER-Tuning
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL + FINER-TuningSize=7B2026.03 | 80.8 | — | — | |
| Qwen2.5-VLSize=7B2026.03 | 74.8 | — | — | |
| InternVL-3.5 + FINER-TuningSize=8B2026.03 | 73.5 | — | — | |
| InternVL-3.5 + FINER-TuningSize=14B2026.03 | 71 | — | — | |
| InternVL-3.5Size=14B2026.03 | 65 | — | — | |
| LLaVA-1.6 + FINER-TuningSize=7B2026.03 | 63.5 | — | — | |
| OmniLMM + RLAIF-VSize=12B2026.03 | 62.4 | — | — | |
| InternVL-3.5Size=8B2026.03 | 62.4 | — | — | |
| Qwen2.5-VL-7B+VRGABackbone=Qwen2.5-VL-7B, Enhancement=VRGA2026.03 | 54.9 | 73.96 | 57.8 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.03 | 50.2 | 66.67 | 59.5 | |
| Qwen2.5-VL-3B+VRGABackbone=Qwen2.5-VL-3B, Enhancement=VRGA2026.03 | 48.8 | 59.03 | 40.5 | |
| Qwen3-VL-30B+VRGABackbone=Qwen3-VL-30B, Enhancement=VRGA2026.03 | 46.5 | 68.05 | 74.4 | |
| Qwen3-VL-30BBackbone=Qwen3-VL-30B2026.03 | 45.4 | 69.2 | 76.2 | |
| Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B2026.03 | 44.5 | 58.87 | 60.1 | |
| LLaVA-1.6Size=7B2026.03 | 44.2 | — | — | |
| Qwen2-VL-7BBackbone=Qwen2-VL-7B2026.03 | 40.6 | 53.81 | 62.2 | |
| Qwen2-VL-7B+VRGABackbone=Qwen2-VL-7B, Enhancement=VRGA2026.03 | 40.4 | 54.47 | 60.8 | |
| OmniLMMSize=12B2026.03 | 39.9 | — | — | |
| Qwen2-VL-7B+ICOTBackbone=Qwen2-VL-7B, Enhancement=ICOT2026.03 | 36.7 | 47.85 | 63.8 | |
| Qwen2-VL-7B+CCOTBackbone=Qwen2-VL-7B, Enhancement=CCOT2026.03 | — | 29.3 | — | |
| Qwen2.5-VL-3B+CCOTBackbone=Qwen2.5-VL-3B, Enhancement=CCOT2026.03 | — | 41.06 | — | |
| Qwen2.5-VL-7B+CCOTBackbone=Qwen2.5-VL-7B, Enhancement=CCOT2026.03 | — | 51.08 | — |