OOD Detection on Texture OOD vs ImageNet-1k ID (test)
22.37FPR@95NegRefine
Evaluation results
| Method | Links | ||
|---|---|---|---|
| NegRefineExtra Resources=Yes, Category=Post-hoc2026.04 | 22.37 | 94.39 | |
| CSPExtra Resources=Yes, Category=Post-hoc2026.04 | 25.52 | 93.86 | |
| TTLExtra Resources=No, Category=Test-time adaptation2026.04 | 26.39 | 94.6 | |
| OODDExtra Resources=No, Category=Test-time adaptation2026.04 | 28.53 | 93.84 | |
| AdaNegExtra Resources=Yes, Category=Test-time adaptation2026.04 | 29.27 | 94.34 | |
| FAExtra Resources=No, Category=Training-based2026.04 | 30.5 | 92.66 | |
| MoFEExtra Resources=No, Category=Training-based2026.04 | 31.47 | 92.15 | |
| AdaNDExtra Resources=No, Category=Test-time adaptation2026.04 | 32.61 | 91.32 | |
| IDLikeExtra Resources=No, Category=Training-based2026.04 | 34.69 | 91.9 | |
| CLIPNExtra Resources=Yes, Category=Post-hoc2026.04 | 41.23 | 90.12 | |
| OSPCoOpExtra Resources=Yes, Category=Training-based2026.04 | 41.26 | 91.13 | |
| LoCoOpExtra Resources=No, Category=Training-based2026.04 | 42.28 | 90.19 | |
| EnergyBackbone=CLIP-L, Training Protocol=Fine-tuned2022.11 | 44.35 | 89.64 | |
| NeglabelExtra Resources=Yes, Category=Post-hoc2026.04 | 45 | 90.02 | |
| APTExtra Resources=Yes, Category=Training-based2026.04 | 45.78 | 88.43 | |
| CMAExtra Resources=Yes, Category=Post-hoc2026.04 | 47.35 | 88.47 | |
| Local-PromptExtra Resources=No, Category=Training-based2026.04 | 48.47 | 88.84 | |
| EnergyBackbone=CLIP-B, Training Protocol=Fine-tuned2022.11 | 51.18 | 88.09 | |
| Fort et al.Backbone=ViT-L, Training Protocol=Fine-tuned2022.11 | 51.38 | 85.54 | |
| Fort et al.Backbone=ViT-B, Training Protocol=Fine-tuned2022.11 | 53.32 | 84.77 | |
| MCMBackbone=CLIP-B, Training Protocol=Zero-shot2022.11 | 57.77 | 86.11 | |
| MCMExtra Resources=No, Category=Post-hoc2026.04 | 57.85 | 86.11 | |
| GL-MCMExtra Resources=No, Category=Post-hoc2026.04 | 58.94 | 83.11 | |
| MSPBackbone=CLIP-L, Training Protocol=Fine-tuned2022.11 | 59.27 | 82.31 | |
| MCMBackbone=CLIP-L, Training Protocol=Zero-shot2022.11 | 59.88 | 84.88 | |
| MOSBackbone=BiT, Training Protocol=Fine-tuned2022.11 | 60.43 | 81.23 | |
| MSPBackbone=CLIP-B, Training Protocol=Fine-tuned2022.11 | 64.96 | 78.16 |