OOD Detection on OpenOOD CIFAR10 Near-OOD
95.72AUROCOnline Prototype Learning
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Online Prototype LearningTraining paradigm=Zero-Shot Training-free Methods2026.05 | 95.72 | 16.38 | |
| TINSBackbone=ViT-B/16, Training Protocol=Zero-shot methods2026.05 | 95.53 | 17.73 | |
| InterNegBackbone=CLIP ViT-B/16, Evaluation Protocol=zero-shot2026.03 | 95.13 | 23.93 | |
| InterNegBackbone=ViT-B/16, Training Protocol=Zero-shot methods2026.05 | 95.13 | 23.93 | |
| PixMix + RotPredTraining Requirement=requiring training (or fine-tuning)2024.10 | 94.86 | — | |
| PixMix + RotPredBackbone=CLIP ViT-B/16, Evaluation Protocol=requires training on ID or extra data2026.03 | 94.86 | — | |
| PixMix + RotPredBackbone=ViT-B/16, Training Protocol=Methods requiring training on ID or extra data2026.05 | 94.86 | — | |
| PixMix + RotPredTraining paradigm=Methods requiring training (or fine-tuning)2026.05 | 94.86 | — | |
| OE + MSPTraining Requirement=requiring training (or fine-tuning)2024.10 | 94.82 | — | |
| OE + MSPBackbone=CLIP ViT-B/16, Evaluation Protocol=requires training on ID or extra data2026.03 | 94.82 | — | |
| OE + MSPBackbone=ViT-B/16, Training Protocol=Methods requiring training on ID or extra data2026.05 | 94.82 | — | |
| OE + MSPTraining paradigm=Methods requiring training (or fine-tuning)2026.05 | 94.82 | — | |
| AdaNegTraining Requirement=Zero-shot Training-free2024.10 | 94.78 | 20.4 | |
| AdaNegBackbone=CLIP ViT-B/16, Evaluation Protocol=zero-shot2026.03 | 94.78 | 20.4 | |
| AdaNegBackbone=ViT-B/16, Training Protocol=Zero-shot methods2026.05 | 94.78 | 20.4 | |
| AdaNeg+NeglabelTraining paradigm=Zero-Shot Training-free Methods2026.05 | 94.78 | 20.4 | |
| NegLabelTraining Requirement=Zero-shot Training-free2024.10 | 94.58 | 28.75 | |
| NegLabelBackbone=CLIP ViT-B/16, Evaluation Protocol=zero-shot2026.03 | 94.58 | 28.75 | |
| NegLabelBackbone=ViT-B/16, Training Protocol=Zero-shot methods2026.05 | 94.58 | 28.75 | |
| NegLabelTraining paradigm=Zero-Shot Training-free Methods2026.05 | 94.58 | 28.75 | |
| PixMix + KNNTraining Requirement=requiring training (or fine-tuning)2024.10 | 93.1 | — | |
| PixMix + KNNBackbone=CLIP ViT-B/16, Evaluation Protocol=requires training on ID or extra data2026.03 | 93.1 | — | |
| PixMix + KNNBackbone=ViT-B/16, Training Protocol=Methods requiring training on ID or extra data2026.05 | 93.1 | — | |
| PixMix + KNNTraining paradigm=Methods requiring training (or fine-tuning)2026.05 | 93.1 | — | |
| MCMTraining Requirement=Zero-shot Training-free2024.10 | 91.92 | 30.86 | |
| MCMBackbone=CLIP ViT-B/16, Evaluation Protocol=zero-shot2026.03 | 91.92 | 30.86 | |
| MCMBackbone=ViT-B/16, Training Protocol=Zero-shot methods2026.05 | 91.92 | 30.86 | |
| MCMTraining paradigm=Zero-Shot Training-free Methods2026.05 | 91.92 | 30.86 | |
| KNNtype=post-processing baseline2026.04 | 90.6 | — | |
| MSPtype=post-processing baseline2026.04 | 88 | — | |
| Energytype=post-processing baseline2026.04 | 87.6 | — | |
| Functional Attribution (CCC)Strategy=online2026.04 | 86.5 | — | |
| Functional Attribution (UMAP)Strategy=offline2026.04 | 85.2 | — | |
| MDStype=post-processing baseline2026.04 | 84.2 | — | |
| ODINtype=post-processing baseline2026.04 | 82.9 | — | |
| SHEtype=post-processing baseline2026.04 | 81.5 | — |