Video Hallucination Evaluation on EventHallusion
84.2Entire ScoreDino-Heal
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Dino-HealBackbone=LLaVA-NeXT2026.01 | 84.2 | 4.7 | 93.1 | 48.9 | — | — | |
| DTRBackbone=Qwen3-VL-4B, Full Name=Decoder-side Temporal Rebalancing2026.04 | 68.42 | 59.07 | 97.06 | 71.15 | — | — | |
| VPSModel=Qwen2.5-VL-7B, J=4, Frames=82025.09 | 65.8 | 91.2 | 56 | 67.5 | — | — | |
| Qwen3-VL-4BBackbone=Qwen3-VL-4B2026.04 | 65.79 | 49.74 | 97.06 | 66.01 | — | — | |
| VCDBackbone=Qwen3-VL-4B, Technique=Visual Contrastive Decoding2026.04 | 65.79 | 49.22 | 98.04 | 66.01 | — | — | |
| DINO-HEALBackbone=Qwen3-VL-4B2026.04 | 64.04 | 47.67 | 97.06 | 64.55 | — | — | |
| VPSModel=Gemma3-12B, J=4, Frames=82025.09 | 62.3 | 83.3 | 82.4 | 77 | — | — | |
| TCDBackbone=Qwen3-VL-4B, Technique=Temporal Contrastive Decoding2026.04 | 62.28 | 51.81 | 97.06 | 66.01 | — | — | |
| VPSModel=Gemma3-12B, J=2, Frames=82025.09 | 61.2 | 82.9 | 81.2 | 76.1 | — | — | |
| VPSModel=Qwen2.5-VL-7B, J=2, Frames=82025.09 | 60.5 | 87.2 | 56 | 65 | — | — | |
| BaselineModel=Gemma3-12B, Frames=82025.09 | 60.5 | 82.4 | 80.3 | 75.3 | — | — | |
| BaselineModel=Qwen2.5-VL-7B, Frames=82025.09 | 57.9 | 84.3 | 48.7 | 60.1 | — | — | |
| SSCDBackbone=LLaVA-NeXT2026.01 | 50.9 | 48.2 | 57.8 | 51.3 | — | — | |
| LLaVA-NeXTBackbone=LLaVA-NeXT2026.01 | 46.5 | 52.3 | 44.1 | 48.7 | — | — | |
| SSCDBackbone=Video-LLaVA2026.01 | 45.6 | 51.3 | 67.6 | 53.8 | — | — | |
| TCDBackbone=LLaVA-NeXT2026.01 | 45.6 | 51.8 | 47.1 | 50.9 | — | — | |
| MotionCDBackbone=LLaVA-NeXT2026.01 | 44.7 | 42.5 | 50 | 45 | — | — | |
| DTRBase Model=InternVL3.5-14B2026.04 | 42.98 | 50.78 | 82.35 | 56.48 | — | — | |
| BaselineModel=InternVL3-8B, Frames=82025.09 | 42.1 | 84.3 | 61.6 | 61.8 | — | — | |
| TCDBase Model=InternVL3.5-14B2026.04 | 41.23 | 50.26 | 79.41 | 55.01 | — | — | |
| VPSModel=InternVL3-8B, J=4, Frames=82025.09 | 41.2 | 84.3 | 63.2 | 62.3 | — | — | |
| VPSModel=InternVL3-8B, J=2, Frames=82025.09 | 40.4 | 85.2 | 62.2 | 61.9 | — | — | |
| InternVL3.5-14BBase Model=InternVL3.5-14B2026.04 | 40.35 | 47.67 | 82.35 | 54.28 | — | — | |
| DINO-HEALBase Model=InternVL3.5-14B2026.04 | 40.35 | 47.15 | 82.35 | 54.03 | — | — | |
| VCDBase Model=InternVL3.5-14B2026.04 | 37.72 | 45.08 | 77.45 | 51.1 | — | — | |
| MotionCDBackbone=Video-LLaVA2026.01 | 32.5 | 48.7 | 33.3 | 40.3 | — | — | |
| Video-LLaVABackbone=Video-LLaVA2026.01 | 31.6 | 57.5 | 43.1 | 46.7 | — | — | |
| TCDBackbone=Video-LLaVA2026.01 | 29.8 | 56.5 | 41.2 | 45.2 | — | — | |
| Dino-HealBackbone=Video-LLaVA2026.01 | 29.8 | 59.6 | 44.1 | 47.4 | — | — | |
| CoF-InternVL2.5-4BParameters=4B2025.05 | — | — | — | — | 71.2 | 64.6 | |
| CoF-InternVL3-8BParameters=8B2025.05 | — | — | — | — | 78.7 | 72.1 | |
| Gemini-1.5-ProAccess=Closed-source2025.05 | — | — | — | — | 80.4 | — | |
| GPT-4oAccess=Closed-source2025.05 | — | — | — | — | 91.9 | — | |
| GPT-4V/4TAccess=Closed-source2025.05 | — | — | — | — | 76.5 | — | |
| InternVL2.5-4BAccess=Open-source, Parameters=4B2025.05 | — | — | — | — | 67.4 | 60.8 | |
| InternVL3-8BAccess=Open-source, Parameters=8B2025.05 | — | — | — | — | 72.1 | 67 | |
| LLaVA-NeXT-Video-7BAccess=Open-source, Parameters=7B2025.05 | — | — | — | — | 64.8 | 50.2 | |
| LLaVA-OneVision-72BAccess=Open-source, Parameters=72B2025.05 | — | — | — | — | 59.5 | 58 | |
| LLaVA-OneVision-7BAccess=Open-source, Parameters=7B2025.05 | — | — | — | — | 60.1 | 53.2 | |
| Qwen2-VL-72BAccess=Open-source, Parameters=72B2025.05 | — | — | — | — | 54.7 | 62.7 | |
| Qwen2-VL-7BAccess=Open-source, Parameters=7B2025.05 | — | — | — | — | 59.3 | 58 |