OOD Detection on ImageNet-1k (ID) vs Places (OOD) 1.0 (test)
96.22AUROCTTL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TTLExtra Resources=No, Category=Test-time adaptation2026.04 | 96.22 | 15.86 | |
| TANLEncoder=VITB/16 CLIP, Protocol=Training-free & test-time adaptation methods2026.03 | 95.87 | 21.9 | |
| Peng et al.Encoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 95.81 | 31.15 | |
| MoFEExtra Resources=No, Category=Training-based2026.04 | 94.69 | 21.32 | |
| AdaNDEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 94.55 | 20.95 | |
| AdaNegEncoder=VITB/16 CLIP, Protocol=Training-free & test-time adaptation methods2026.03 | 94.55 | 34.34 | |
| OSPCoOpExtra Resources=Yes, Category=Training-based2026.04 | 94.01 | 25.74 | |
| AdaNegExtra Resources=Yes, Category=Test-time adaptation2026.04 | 93.69 | 35.16 | |
| CLIPScopeEncoder=VITB/16 CLIP, Protocol=Training-free & test-time adaptation methods2026.03 | 93.54 | 28.45 | |
| SynOODEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 93.37 | 12.12 | |
| NegPromptEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 93.34 | 27.6 | |
| CMA [37]Encoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 93.17 | 29.86 | |
| CMAExtra Resources=Yes, Category=Post-hoc2026.04 | 93.17 | 29.86 | |
| CMA [30]Encoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 93.11 | 27.65 | |
| EOEEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 92.95 | 30.16 | |
| CSPEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 92.9 | 29.32 | |
| CSPExtra Resources=Yes, Category=Post-hoc2026.04 | 92.9 | 29.32 | |
| APTExtra Resources=Yes, Category=Training-based2026.04 | 92.84 | 28.54 | |
| CLIPNExtra Resources=Yes, Category=Post-hoc2026.04 | 92.62 | 32.26 | |
| FAExtra Resources=No, Category=Training-based2026.04 | 92.54 | 30.3 | |
| OODDEncoder=VITB/16 CLIP, Protocol=Training-free & test-time adaptation methods2026.03 | 92.51 | 30.68 | |
| Local-PromptExtra Resources=No, Category=Training-based2026.04 | 92.5 | 32.43 | |
| FAGLEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 92.44 | 31.09 | |
| CLIPNEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 92.28 | 33.45 | |
| AdaNDExtra Resources=No, Category=Test-time adaptation2026.04 | 92.27 | 31.41 | |
| LAPTEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 92.01 | 33.01 | |
| LoCoOpEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 91.98 | 32.87 | |
| LoCoOpExtra Resources=No, Category=Training-based2026.04 | 91.98 | 32.87 | |
| DualCnstEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 91.81 | 31.77 | |
| NegLabelEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 91.64 | 35.59 | |
| NeglabelExtra Resources=Yes, Category=Post-hoc2026.04 | 91.56 | 36.48 | |
| EnergyEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 91.41 | 39.87 | |
| LSNEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 91.25 | 34.48 | |
| VOSEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 91.23 | 38.39 | |
| KNNEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 91.02 | 39.61 | |
| GL-MCMExtra Resources=No, Category=Post-hoc2026.04 | 90.37 | 37.07 | |
| MCMEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 90.31 | 46.2 | |
| CoVerEncoder=VITB/16 CLIP, Protocol=Training-free & non-adaptive methods2026.03 | 90.27 | 40.71 | |
| NegRefineExtra Resources=Yes, Category=Post-hoc2026.04 | 89.91 | 40.39 | |
| MCMExtra Resources=No, Category=Post-hoc2026.04 | 89.77 | 44.71 | |
| NPOSEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 89.44 | 45.27 | |
| OODDExtra Resources=No, Category=Test-time adaptation2026.04 | 88.13 | 41.91 | |
| TagOODEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 87.81 | 40.4 | |
| IDLikeExtra Resources=No, Category=Training-based2026.04 | 85.86 | 56.63 | |
| ODINEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 85.54 | 55.06 | |
| ViMEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 83.75 | 60.67 | |
| ZOCEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 83.39 | 73.06 | |
| MSPEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 79.67 | 74.41 | |
| GradNormEncoder=VITB/16 CLIP, Protocol=Methods requiring training (or fine-tuning)2026.03 | 73.7 | 80.41 |