Image Quality Assessment on KonIQ (test)
0.941SROCCDeQA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeQAModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.941 | 0.953 | — | |
| DeQACategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.941 | 0.953 | — | |
| Q-AlignModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.94 | 0.941 | — | |
| Q-AlignCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.94 | 0.941 | — | |
| MUSIQModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.929 | 0.924 | — | |
| MUSIQCategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.929 | 0.924 | — | |
| OursCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.92 | 0.93 | — | |
| Autoencoder-likeModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.912 | 0.926 | — | |
| C2ScoreModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.91 | 0.923 | — | |
| C2ScoreCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.91 | 0.923 | — | |
| Chain-of-ThoughtModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.907 | 0.92 | — | |
| Self-ConsistencyModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.905 | 0.917 | — | |
| CLIP-IQA+Model category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.895 | 0.909 | — | |
| Q-Insight-ScoreModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.895 | 0.918 | — | |
| CLIP-IQA+Category=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.895 | 0.909 | — | |
| Q-Insight-ScoreCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.895 | 0.918 | — | |
| Self-ConsistencyInput Condition=Text-Only2026.01 | 0.879 | 0.9 | — | |
| Self-ConsistencyInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.879 | 0.898 | — | |
| DBCNNModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.875 | 0.884 | — | |
| DBCNNCategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.875 | 0.884 | — | |
| Autoencoder-likeInput Condition=Text-Only2026.01 | 0.861 | 0.877 | — | |
| NIMAModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.859 | 0.896 | — | |
| NIMACategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.859 | 0.896 | — | |
| OursCategory=Caption-Only Conditions, Condition=Caption-Only2025.12 | 0.855 | 0.871 | — | |
| Autoencoder-likeInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.847 | 0.867 | — | |
| MANIQAModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.834 | 0.849 | — | |
| MANIQACategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.834 | 0.849 | — | |
| Q-Insight-ScoreInput Condition=Text-Only2026.01 | 0.827 | 0.859 | — | |
| Q-Insight-ScoreInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.825 | 0.856 | — | |
| Chain-of-ThoughtInput Condition=Text-Only2026.01 | 0.819 | 0.851 | — | |
| Chain-of-ThoughtInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.818 | 0.851 | — | |
| Q-Insight-ScoreCategory=Caption-Only Conditions, Condition=Caption-Only2025.12 | 0.818 | 0.841 | — | |
| SaTQATraining on=CLIVE2024.05 | 0.788 | — | — | |
| Ours (student)Training on=CLIVE2024.05 | 0.781 | — | — | |
| HyperIQATraining Dataset=CLIVE2021.10 | 0.772 | — | — | |
| HyperIQATraining on=CLIVE2024.05 | 0.772 | — | — | |
| Ours (teacher)Training on=LIVEFB2024.05 | 0.771 | — | — | |
| Ours (student)Training on=LIVEFB2024.05 | 0.767 | — | — | |
| Ours (teacher)Training on=CLIVE2024.05 | 0.766 | — | — | |
| LoDaTraining on=LIVEFB2024.05 | 0.763 | — | — | |
| HyperIQATraining on=LIVEFB2024.05 | 0.758 | — | — | |
| PQRTraining Dataset=CLIVE2021.10 | 0.757 | — | — | |
| P2P-BMTraining on=LIVEFB2024.05 | 0.755 | — | — | |
| DB-CNNTraining Dataset=CLIVE2021.10 | 0.754 | — | — | |
| DBCNNTraining on=CLIVE2024.05 | 0.754 | — | — | |
| LoDaTraining on=CLIVE2024.05 | 0.745 | — | — | |
| P2P-BMTraining on=CLIVE2024.05 | 0.74 | — | — | |
| TReSTraining on=CLIVE2024.05 | 0.733 | — | — | |
| DBCNNTraining on=LIVEFB2024.05 | 0.716 | — | — | |
| TReSTraining on=LIVEFB2024.05 | 0.713 | — | — | |
| CONTRIQUETraining Dataset=CLIVE2021.10 | 0.676 | — | — | |
| NIQECategory=Hand-Crafted Models, Condition=Image-conditioned2025.12 | 0.53 | 0.533 | — | |
| BRISQUECategory=Hand-Crafted Models, Condition=Image-conditioned2025.12 | 0.226 | 0.225 | — | |
| Active EloHuman comparisons=6002026.06 | — | — | 0.598 | |
| Active Elo (same budget)Human comparisons=6002026.06 | — | — | 0.583 | |
| PairS-MSHuman comparisons=6002026.06 | — | — | 0.338 | |
| RAIR-LaplacianHuman comparisons=6002026.06 | — | — | 0.653 | |
| Random EloHuman comparisons=6002026.06 | — | — | 0.655 | |
| SGSHuman comparisons=600, Inferred comparisons=5352026.06 | — | — | 0.658 |