Video Quality Assessment on LSVQ 1080p
0.832SRCCSoft Ranking (Stage 1)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Soft Ranking (Stage 1)Training Stage=Stage 12025.05 | 0.832 | 0.857 | |
| VersusQMethod Category=Reasoning or reinforcement-learning based LMM evaluators2026.05 | 0.824 | 0.852 | |
| Q-ALIGN (1fps) + FAST-VQATraining Set=LSVQtrain, IQA Pre-training=false, fps=12023.12 | 0.818 | 0.85 | |
| KVQGFlops=3532025.03 | 0.814 | 0.846 | |
| qstInference Time (Sec)=0.159, Testing Protocol=In-dataset Testing2024.02 | 0.809 | 0.844 | |
| VQ-Jarvis2026.03 | 0.809 | 0.84 | |
| LMM-PVQA (Soft Ranking Stage 3)Training data=PLVD-P1/ P2/ P3 (700k), Human Label=No, Setting=Zero-shot2025.05 | 0.806 | 0.835 | |
| LMM-PVQA (Soft Ranking Stage 1)Training data=PLVD-P1 (500k), Human Label=No, Setting=Zero-shot2025.05 | 0.803 | 0.83 | |
| LMM-PVQA (Soft Ranking Stage 2)Training data=PLVD-P1/ P2 (600k), Human Label=No, Setting=Zero-shot2025.05 | 0.802 | 0.83 | |
| LMM-PVQA (Hard Ranking)Training data=PLVD-P1 (500k), Human Label=No, Setting=Zero-shot2025.05 | 0.799 | 0.817 | |
| VQAThinkerMethod Category=Reasoning or reinforcement-learning based LMM evaluators2026.05 | 0.798 | 0.834 | |
| Q-ALIGN (1fps)Training Set=LSVQtrain, IQA Pre-training=false, fps=12023.12 | 0.797 | 0.83 | |
| Q-AlignModel Type=Specialized VQA Model2026.02 | 0.797 | 0.83 | |
| Q-AlignEvaluation Protocol=Intra-dataset2024.07 | 0.797 | 0.83 | |
| qtInference Time (Sec)=0.159, Testing Protocol=In-dataset Testing2024.02 | 0.796 | 0.831 | |
| DOVERTraining Set=LSVQtrain, IQA Pre-training=false, Components=aesthetic branch + FAST-VQA2023.12 | 0.795 | 0.83 | |
| DOVERInference Time (Sec)=0.047, Testing Protocol=In-dataset Testing2024.02 | 0.795 | 0.83 | |
| VQA2-ScorerMethod Category=Supervised VQA/IQA models2026.05 | 0.794 | 0.821 | |
| MinimalisticVQA(IX)Training data=LSVQ [70], Human Label=Yes, Setting=Zero-shot2025.05 | 0.792 | 0.828 | |
| MinimalisticVQAMethod Category=Supervised VQA/IQA models2026.05 | 0.792 | 0.828 | |
| Modular-VQA2026.03 | 0.791 | 0.844 | |
| KSVQE2024.02 | 0.79 | 0.823 | |
| Soft Ranking (Base)Training Stage=Base2025.05 | 0.79 | 0.814 | |
| DOVER2026.03 | 0.787 | 0.828 | |
| VQ-Insight2026.03 | 0.786 | 0.823 | |
| VQ-InsightMethod Category=Reasoning or reinforcement-learning based LMM evaluators2026.05 | 0.786 | 0.823 | |
| CLIPVQAGFlops=6582025.03 | 0.785 | 0.833 | |
| QPT V2Evaluation Protocol=Intra-dataset2024.07 | 0.785 | 0.822 | |
| DOVER2025.05 | 0.782 | 0.813 | |
| DOVERTraining data=LSVQ [70], Human Label=Yes, Setting=Zero-shot2025.05 | 0.782 | 0.813 | |
| DOVERMethod Category=Supervised VQA/IQA models2026.05 | 0.782 | 0.813 | |
| FAST-VQA2025.05 | 0.781 | 0.813 | |
| MinimalisticVQA2025.05 | 0.781 | 0.82 | |
| FAST-VQATraining data=LSVQ [70], Human Label=Yes, Setting=Zero-shot2025.05 | 0.781 | 0.813 | |
| FAST-VQAMethod Category=Supervised VQA/IQA models2026.05 | 0.781 | 0.813 | |
| FAST-VQATraining Set=LSVQtrain, IQA Pre-training=false2023.12 | 0.779 | 0.814 | |
| FastVQA2024.02 | 0.779 | 0.814 | |
| FastVQAInference Time (Sec)=0.045, Testing Protocol=In-dataset Testing2024.02 | 0.779 | 0.814 | |
| Fast-VQAGFlops=279.12025.03 | 0.779 | 0.814 | |
| FastVQAEvaluation Protocol=Intra-dataset2024.07 | 0.779 | 0.814 | |
| Doveraesthetic branch=false2024.02 | 0.778 | 0.812 | |
| BVQATraining Set=LSVQtrain, IQA Pre-training=true2023.12 | 0.772 | 0.788 | |
| Li22Inference Time (Sec)=27.632, Testing Protocol=In-dataset Testing2024.02 | 0.772 | 0.788 | |
| Faster-VQAGFlops=69.82025.03 | 0.772 | 0.811 | |
| BVQAEvaluation Protocol=Intra-dataset2024.07 | 0.772 | 0.788 | |
| Li et al.GFlops=1125372025.03 | 0.771 | 0.782 | |
| ASOTraining Strategy=Analytic Score Optimization2026.02 | 0.771 | 0.824 | |
| Minimalist-VQA2026.03 | 0.769 | 0.818 | |
| SFTTraining Strategy=Supervised Fine-Tuning2026.02 | 0.765 | 0.737 | |
| Fast-VQA2026.03 | 0.765 | 0.81 | |
| SimpleVQATraining Set=LSVQtrain, IQA Pre-training=true2023.12 | 0.764 | 0.803 | |
| SimpleVQA2024.02 | 0.764 | 0.803 | |
| qsInference Time (Sec)=0.159, Testing Protocol=In-dataset Testing2024.02 | 0.764 | 0.808 | |
| SimpleVQAEvaluation Protocol=Intra-dataset2024.07 | 0.764 | 0.803 | |
| SimpleVQA2025.05 | 0.764 | 0.803 | |
| Q-AlignTraining data=fused [11, 17, 28, 40, 70], Human Label=Yes, Setting=Zero-shot2025.05 | 0.761 | 0.822 | |
| Q-AlignMethod Category=Supervised VQA/IQA models2026.05 | 0.761 | 0.822 | |
| VQA^22026.03 | 0.76 | 0.819 | |
| Q-Align2026.03 | 0.758 | 0.833 | |
| SimpleVQAMethod Category=Supervised VQA/IQA models2026.05 | 0.756 | 0.801 | |
| qbInference Time (Sec)=0.159, Testing Protocol=In-dataset Testing2024.02 | 0.754 | 0.802 | |
| SimpleVQAInference Time (Sec)=0.714, Testing Protocol=In-dataset Testing2024.02 | 0.75 | 0.793 | |
| MinimalisticVQA(VII)Training data=LSVQ [70], Human Label=Yes, Setting=Zero-shot2025.05 | 0.74 | 0.784 | |
| DisCoVQATraining Set=LSVQtrain, IQA Pre-training=false2023.12 | 0.734 | 0.772 | |
| VisualQuality-R1Method Category=Reasoning or reinforcement-learning based LMM evaluators2026.05 | 0.716 | 0.744 | |
| PVQTraining Set=LSVQtrain, IQA Pre-training=true2023.12 | 0.711 | 0.739 | |
| PVQ2024.02 | 0.711 | 0.739 | |
| PVQw/patchGFlops=585012025.03 | 0.711 | 0.739 | |
| PVQEvaluation Protocol=Intra-dataset, Patch usage=w/ patch2024.07 | 0.711 | 0.739 | |
| PatchVQ2025.05 | 0.711 | 0.739 | |
| VideoLLaMA3 (Qwen2.5-7B)S2I-Tuned=Yes2025.06 | 0.705 | 0.714 | |
| PVQwo/patchGFlops=585012025.03 | 0.686 | 0.708 | |
| PVQEvaluation Protocol=Intra-dataset, Patch usage=w/o patch2024.07 | 0.686 | 0.708 | |
| Gemini-2.5ProModel Type=VLM-API2026.02 | 0.682 | 0.691 | |
| Video-LLaVA (Vicuna-v1.5-7B)S2I-Tuned=Yes2025.06 | 0.68 | 0.653 | |
| VSFATraining Set=LSVQtrain, IQA Pre-training=false2023.12 | 0.675 | 0.704 | |
| VSFA2024.02 | 0.675 | 0.704 | |
| VSFAInference Time (Sec)=11.109, Testing Protocol=In-dataset Testing2024.02 | 0.675 | 0.704 | |
| VSFAGFlops=409192025.03 | 0.675 | 0.704 | |
| VSFAEvaluation Protocol=Intra-dataset2024.07 | 0.675 | 0.704 | |
| LLaVA-Next-Video (Mistral-7B)S2I-Tuned=Yes2025.06 | 0.675 | 0.705 | |
| VSFA2025.05 | 0.675 | 0.704 | |
| LLaVA-OneVision (Vicuna-v1.1-7B)S2I-Tuned=Yes2025.06 | 0.671 | 0.634 | |
| LLaVA-Video (Qwen2-7B)S2I-Tuned=Yes2025.06 | 0.667 | 0.652 | |
| GPT-4.1Model Type=VLM-API2026.02 | 0.651 | 0.677 | |
| Q-Instruct2026.03 | 0.644 | 0.64 | |
| Qwen2.5-VLModel Type=Open-source VLM2026.02 | 0.624 | 0.61 | |
| InternVL-3.5Model Type=Open-source VLM2026.02 | 0.61 | 0.595 | |
| Q-InsightMethod Category=Reasoning or reinforcement-learning based LMM evaluators2026.05 | 0.601 | 0.648 | |
| TLVQMTraining Set=LSVQtrain, IQA Pre-training=false2023.12 | 0.589 | 0.616 | |
| TLVQM2024.02 | 0.589 | 0.616 | |
| TLVQM2025.03 | 0.589 | 0.616 | |
| TLVQMEvaluation Protocol=Intra-dataset2024.07 | 0.589 | 0.616 | |
| TLVQM2025.05 | 0.589 | 0.616 | |
| InternVL-Chat (Vicuna-7B)S2I-Tuned=Yes2025.06 | 0.582 | 0.59 | |
| CLIP-IQAMethod Category=Training-free / unsupervised quality models2026.05 | 0.553 | 0.505 | |
| VIDEVALTraining Set=LSVQtrain, IQA Pre-training=false2023.12 | 0.545 | 0.554 | |
| VIDEVAL2024.02 | 0.545 | 0.554 | |
| VIDEVAL2025.03 | 0.545 | 0.554 | |
| VIDEVALEvaluation Protocol=Intra-dataset2024.07 | 0.545 | 0.554 |