Compositional Zero-Shot Learning on UT-Zappos Closed World
60.5HMCGE
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CGEadopting_CLIP=false2024.02 | 60.5 | 33.5 | 64.5 | 71.5 | — | — | |
| CGECLIP-based=false2023.05 | 60.5 | 33.5 | 64.5 | 71.5 | — | — | |
| H2EMSetting=Closed-World2025.12 | 59.8 | 46.6 | — | — | 70.4 | 74.5 | |
| Troika + FlowComposerbaseline=Troika2026.03 | 58.6 | 46.8 | — | — | 71.1 | 74.9 | |
| LOGICZSLSetting=Closed-World2025.12 | 57.8 | 45.8 | — | — | 69.6 | 74.9 | |
| LOGICZSLleveraging LLM knowledge=true2026.03 | 57.8 | 45.8 | — | — | 69.6 | 74.9 | |
| CAILACLIP-based=true, Backbone=ViT-L/142023.05 | 57 | 44.1 | 67.8 | 74 | — | — | |
| CAILASetting=Closed-World, Reproduction=Official Code2025.12 | 56 | 42.5 | — | — | 66.8 | 72.5 | |
| Troikaimplementation source=re-implemented2026.03 | 55.4 | 41.8 | — | — | 66.3 | 73.4 | |
| PLOSetting=Closed-World2025.12 | 55.3 | 43.6 | — | — | 70.3 | 75.8 | |
| PLOleveraging LLM knowledge=true2026.03 | 55.3 | 43.6 | — | — | 70.3 | 75.8 | |
| TroikaSetting=Closed-World2025.12 | 54.6 | 41.7 | — | — | 66.8 | 73.8 | |
| IMAX2026.03 | 54.2 | 40.6 | — | — | 69.3 | 70.7 | |
| CDS-CZSLadopting_CLIP=true2024.02 | 52.7 | 39.5 | 63.9 | 74.8 | — | — | |
| CDS-CZSLSetting=Closed-World2025.12 | 52.7 | 39.5 | — | — | 63.9 | 74.8 | |
| CDS-CZSL2026.03 | 52.7 | 39.5 | — | — | 63.9 | 74.8 | |
| PLIDSetting=Closed-World2025.12 | 52.4 | 38.7 | — | — | 67.3 | 68.8 | |
| PLIDleveraging LLM knowledge=true2026.03 | 52.4 | 38.7 | — | — | 67.3 | 68.8 | |
| DLM2026.03 | 52 | 39.6 | — | — | 67.1 | 72.5 | |
| CSP + FlowComposerbaseline=CSP2026.03 | 51.2 | 37.8 | — | — | 66.6 | 68.2 | |
| ADEadopting_CLIP=false2024.02 | 51.1 | 35.1 | 63 | 64.3 | — | — | |
| ProtoPropCLIP-based=false2023.05 | 50.2 | 34.7 | 62.1 | 65.7 | — | — | |
| CLIP-based Co-CGEadopting_CLIP=true2024.02 | 49.7 | 36.3 | 63.4 | 71.3 | — | — | |
| Co-CGECLIP-based=true, Backbone=ViT-L/142023.05 | 49.7 | 36.3 | 63.4 | 71.3 | — | — | |
| CAPECLIP-based=false2023.05 | 49.5 | 35.2 | 62.3 | 68.5 | — | — | |
| GIPCOLSetting=Closed-World2025.12 | 48.8 | 36.2 | — | — | 65 | 68.5 | |
| HPLadopting_CLIP=true2024.02 | 48.2 | 35 | 63 | 68.8 | — | — | |
| Co-CGEadopting_CLIP=false2024.02 | 48.1 | 33.9 | 62.3 | 66.3 | — | — | |
| Co-CGECLIP-based=false2023.05 | 48.1 | 33.9 | 62.3 | 66.3 | — | — | |
| SCENadopting_CLIP=false2024.02 | 47.8 | 32 | 63.5 | 63.1 | — | — | |
| SCENCLIP-based=false2023.05 | 47.8 | 32 | 63.5 | 63.1 | — | — | |
| CANetadopting_CLIP=false2024.02 | 47.3 | 33.1 | 61 | 66.3 | — | — | |
| DFSPadopting_CLIP=true2024.02 | 47.2 | 36 | 66.7 | 71.7 | — | — | |
| DFSP(t2i)Setting=Closed-World2025.12 | 47.2 | 36 | — | — | 66.7 | 71.7 | |
| DFSPCLIP-based=true, Backbone=ViT-L/142023.05 | 47.2 | 36 | 66.7 | 71.7 | — | — | |
| DFSP(t2i)2026.03 | 47.2 | 36 | — | — | 66.7 | 71.7 | |
| DFSP(BiF)Setting=Closed-World2025.12 | 47.1 | 33.5 | — | — | 63.3 | 69.2 | |
| DFSP(BiF)2026.03 | 47.1 | 33.5 | — | — | 63.3 | 69.2 | |
| CSPimplementation source=re-implemented2026.03 | 47 | 32.4 | — | — | 63.4 | 65.3 | |
| CSPadopting_CLIP=true2024.02 | 46.6 | 33 | 64.2 | 66.2 | — | — | |
| CSPSetting=Closed-World2025.12 | 46.6 | 33 | — | — | 64.2 | 66.2 | |
| CSPCLIP-based=true, Backbone=ViT-L/142023.05 | 46.6 | 33 | 64.2 | 66.2 | — | — | |
| Co-CGESetting=Closed-World2025.12 | 45.3 | 28.4 | — | — | 59.9 | 56.2 | |
| DFSP(i2t)Setting=Closed-World2025.12 | 45.1 | 32.1 | — | — | 64.2 | 66.4 | |
| DFSP(i2t)2026.03 | 45.1 | 32.1 | — | — | 64.2 | 66.4 | |
| OADisCLIP-based=false2023.05 | 44.4 | 30 | 59.5 | 65.5 | — | — | |
| CompCosadopting_CLIP=false2024.02 | 43.1 | 28.7 | 59.8 | 62.5 | — | — | |
| CompCosCLIP-based=false2023.05 | 43.1 | 28.7 | 59.8 | 62.5 | — | — | |
| CoOPadopting_CLIP=true2024.02 | 34.6 | 18.8 | 52.1 | 49.3 | — | — | |
| CoOpSetting=Closed-World2025.12 | 34.6 | 18.8 | — | — | 52.1 | 49.3 | |
| CoOpCLIP-based=true, Backbone=ViT-L/142023.05 | 34.6 | 18.8 | 52.1 | 49.3 | — | — | |
| CoOP2026.03 | 34.6 | 18.8 | — | — | 52.1 | 49.3 | |
| ProDASetting=Closed-World2025.12 | 34.3 | 18.4 | — | — | 63.9 | 34.6 | |
| CLIPadopting_CLIP=true2024.02 | 15.6 | 5 | 15.8 | 49.1 | — | — | |
| CLIPSetting=Closed-World2025.12 | 15.6 | 5 | — | — | 15.8 | 49.1 | |
| CLIP-ZSCLIP-based=true, Backbone=ViT-L/142023.05 | 15.6 | 5 | 15.8 | 49.1 | — | — | |
| CLIP2026.03 | 15.6 | 5 | — | — | 15.8 | 49.1 |