Hallucination on HallusionBench
69.2HallusionBench ScoreAOD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AODModel=Qwen2.5-VL-7B2026.05 | 69.2 | — | — | |
| PruneHalModel=Qwen2.5-VL-7B2026.05 | 67 | — | — | |
| AODModel=InternVL3-8B2026.05 | 66.5 | — | — | |
| TruthPrIntModel=Qwen2.5-VL-7B2026.05 | 66.2 | — | — | |
| ASDModel=Qwen2.5-VL-7B2026.05 | 65.8 | — | — | |
| VASparseModel=Qwen2.5-VL-7B2026.05 | 65.1 | — | — | |
| PruneHalModel=InternVL3-8B2026.05 | 65.1 | — | — | |
| VCDModel=Qwen2.5-VL-7B2026.05 | 64.5 | — | — | |
| AODModel=LLaVA-1.5-7B2026.05 | 64.3 | — | — | |
| BaseModel=Qwen2.5-VL-7B2026.05 | 63.7 | — | — | |
| TruthPrIntModel=InternVL3-8B2026.05 | 63.4 | — | — | |
| TruthPrIntModel=LLaVA-1.5-7B2026.05 | 63.1 | — | — | |
| ASDModel=InternVL3-8B2026.05 | 63.1 | — | — | |
| PruneHalModel=LLaVA-1.5-7B2026.05 | 62.7 | — | — | |
| ASDModel=LLaVA-1.5-7B2026.05 | 62.5 | — | — | |
| VCDModel=InternVL3-8B2026.05 | 61.9 | — | — | |
| VASparseModel=LLaVA-1.5-7B2026.05 | 61.8 | — | — | |
| VCDModel=LLaVA-1.5-7B2026.05 | 61.2 | — | — | |
| VASparseModel=InternVL3-8B2026.05 | 61.2 | — | — | |
| BaseModel=InternVL3-8B2026.05 | 60.7 | — | — | |
| BaseModel=LLaVA-1.5-7B2026.05 | 58.8 | — | — | |
| BAGEL+Ours# Params=7B, Model Type=Unified model2026.05 | 55.1 | — | — | |
| InternVL2.5# Params=7B, Model Type=Understanding Only2026.05 | 52.9 | — | — | |
| BAGEL# Params=7B, Model Type=Unified model2026.05 | 50.9 | — | — | |
| Qwen2.5-VL# Params=3B, Model Type=Understanding Only2026.05 | 46.3 | — | — | |
| InternVL2.5# Params=4B, Model Type=Understanding Only2026.05 | 46.3 | — | — | |
| Claude4-Sonnet2026.02 | — | 54.5 | 64.1 | |
| DPEBackbone=Qwen3-VL-8B-Instruct, Iteration=32026.02 | — | 74.13 | 64.39 | |
| GPT-4o2026.02 | — | 67.5 | 56.1 | |
| GPT5-Mini2026.02 | — | 55.9 | 53.8 | |
| LoMoModel=LLaVA-OV1.5-8B, Evaluation Protocol=Standard Evaluation2026.05 | — | — | 42.82 | |
| LoMoModel=Qwen3.5-9B, Evaluation Protocol=Standard Evaluation2026.05 | — | — | 52.99 | |
| LoMoModel=LLaVA-OV1.5-8B, Evaluation Protocol=Rendered Evaluation2026.05 | — | — | 61.65 | |
| LoMoModel=Qwen3.5-9B, Evaluation Protocol=Rendered Evaluation2026.05 | — | — | 47.05 | |
| MiMo-VLSize=7B2026.06 | — | — | 67.7 | |
| MiMo-VL + LOCUSSize=7B2026.06 | — | — | 72 | |
| Qwen2.5-VLSize=7B2026.06 | — | — | 68 | |
| Qwen2.5-VL + LOCUSSize=7B2026.06 | — | — | 70.3 | |
| Qwen2.5-VL-72BParameters=72B2026.02 | — | 72.4 | 61.9 | |
| Qwen3-VLSize=4B2026.06 | — | — | 75.4 | |
| Qwen3-VL + LOCUSSize=4B2026.06 | — | — | 75.6 | |
| Standard SFTModel=LLaVA-OV1.5-8B, Evaluation Protocol=Standard Evaluation2026.05 | — | — | 40.35 | |
| Standard SFTModel=Qwen3.5-9B, Evaluation Protocol=Standard Evaluation2026.05 | — | — | 51.53 | |
| Standard SFTModel=LLaVA-OV1.5-8B, Evaluation Protocol=Rendered Evaluation2026.05 | — | — | 41.36 | |
| Standard SFTModel=Qwen3.5-9B, Evaluation Protocol=Rendered Evaluation2026.05 | — | — | 43.37 |