Image Classification on LADV-3D (OOD)
68.33AccuracyPCBM+ + DDO
Evaluation results
| Method | Links | |
|---|---|---|
| PCBM+ + DDOBackbone=CLIP ViT-L/14, Concept=ConceptNet2025.03 | 68.33 | |
| PCBM†Concept=ConceptNet, Method=+DDO, Backbone=CLIP2025.03 | 68.33 | |
| LaBO + DDOBackbone=CLIP ViT-L/14, Concept=LLM2025.03 | 68.01 | |
| LaBO‡Concept=LLM, Method=+DDO2025.03 | 68.01 | |
| CLIP ZSBackbone=CLIP ViT-L/14, Concept=none2025.03 | 66.29 | |
| CLIP ZS2025.03 | 66.29 | |
| PCBM+ baselineBackbone=CLIP ViT-L/14, Concept=ConceptNet2025.03 | 65.88 | |
| PCBM†Concept=ConceptNet, Method=baseline, Backbone=CLIP2025.03 | 65.88 | |
| CLIP LPBackbone=CLIP ViT-L/14, Concept=none2025.03 | 63.81 | |
| CLIP LP2025.03 | 63.81 | |
| CLIP-CBM + DDOBackbone=CLIP ViT-L/14, Concept=human2025.03 | 63.51 | |
| CLIP-CBMConcept=human, Method=+DDO2025.03 | 63.51 | |
| LaBO baselineBackbone=CLIP ViT-L/14, Concept=LLM2025.03 | 63.17 | |
| LaBO‡Concept=LLM, Method=baseline2025.03 | 63.17 | |
| CLIP-CBM baselineBackbone=CLIP ViT-L/14, Concept=human2025.03 | 60.64 | |
| CLIP-CBMConcept=human, Method=baseline2025.03 | 60.64 | |
| CLIP LPBackbone=CLIP ViT-B/32, Concept=none2025.03 | 60.49 | |
| LaBO + DDOBackbone=CLIP ViT-B/32, Concept=LLM2025.03 | 58.88 | |
| PCBM+ + DDOBackbone=CLIP ViT-B/32, Concept=ConceptNet2025.03 | 57.18 | |
| PCBM+ baselineBackbone=CLIP ViT-B/32, Concept=ConceptNet2025.03 | 56.87 | |
| LaBO baselineBackbone=CLIP ViT-B/32, Concept=LLM2025.03 | 56 | |
| CLIP-CBM + DDOBackbone=CLIP ViT-B/32, Concept=human2025.03 | 55.42 | |
| CLIP-CBM baselineBackbone=CLIP ViT-B/32, Concept=human2025.03 | 53.25 | |
| CLIP ZSBackbone=CLIP ViT-B/32, Concept=none2025.03 | 50.07 |