Compositional Zero-Shot Learning on UT-Zappos open world
54.5HMH2EM
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| H2EMSetting=Open-World2025.12 | 54.5 | 38.8 | — | — | 70.4 | 62.3 | |
| Troika + FlowComposerbaseline=Troika2026.03 | 51 | 35.5 | — | — | 70.1 | 61.2 | |
| LOGICZSLSetting=Open-World2025.12 | 50.8 | 36.2 | — | — | 69.6 | 63.7 | |
| LOGICZSLleveraging LLM knowledge=true2026.03 | 50.8 | 36.2 | — | — | 69.6 | 63.7 | |
| CAILACLIP-based=true, Visual Backbone=ViT-L/142023.05 | 49.4 | 32.8 | 67.8 | 59.7 | — | — | |
| CDS-CZSLSetting=Open-World2025.12 | 48.2 | 32.3 | — | — | 64.7 | 61.3 | |
| CDS-CZSL2026.03 | 48.2 | 32.3 | — | — | 64.7 | 61.3 | |
| TroikaSetting=Open-World2025.12 | 47.8 | 33 | — | — | 66.4 | 61.2 | |
| PLOSetting=Open-World2025.12 | 47.8 | 33.1 | — | — | 68 | 63.5 | |
| PLOleveraging LLM knowledge=true2026.03 | 47.8 | 33.1 | — | — | 68 | 63.5 | |
| IMAX2026.03 | 47.5 | 32.3 | — | — | 68.4 | 57.3 | |
| PLIDSetting=Open-World2025.12 | 46.6 | 30.8 | — | — | 67.6 | 55.5 | |
| PLIDleveraging LLM knowledge=true2026.03 | 46.6 | 30.8 | — | — | 67.6 | 55.5 | |
| Troikaimplementation source=re-implemented2026.03 | 46.6 | 32.7 | — | — | 65.8 | 61 | |
| Co-CGESetting=Open-World2025.12 | 45.3 | 28.4 | — | — | 59.9 | 56.2 | |
| Co-CGECLIP-based=true, Visual Backbone=ViT-L/14, mode=CLIP features (best reported)2023.05 | 45.3 | 28.4 | 59.9 | 56.2 | — | — | |
| DFSP(i2t)Setting=Open-World2025.12 | 45.1 | 32.1 | — | — | 64.2 | 66.4 | |
| CAILASetting=Open-World, Reproduction=Official Code2025.12 | 44.8 | 29.9 | — | — | 65.1 | 59.6 | |
| DFSP(t2i)Setting=Open-World2025.12 | 44 | 30.3 | — | — | 66.8 | 60 | |
| DFSPCLIP-based=true, Visual Backbone=ViT-L/142023.05 | 44 | 30.3 | 66.8 | 60 | — | — | |
| DFSP(t2i)2026.03 | 44 | 30.3 | — | — | 66.8 | 60 | |
| DFSP(BiF)Setting=Open-World2025.12 | 42.7 | 27.6 | — | — | 63.5 | 57.2 | |
| DFSP(BiF)2026.03 | 42.7 | 27.6 | — | — | 63.5 | 57.2 | |
| KG-SPCLIP-based=false2023.05 | 42.3 | 26.5 | 61.8 | 52.1 | — | — | |
| KG-SPsource=paper2025.12 | 42.3 | 26.5 | 61.8 | 52.1 | — | — | |
| DFSP(i2t)2026.03 | 41.2 | 26.4 | — | — | 64.3 | 53.8 | |
| CSP + FlowComposerbaseline=CSP2026.03 | 41.2 | 25.3 | — | — | 66.4 | 46.2 | |
| KG-SPsource=Implementation2025.12 | 40.9 | 25.7 | 62.6 | 51.5 | — | — | |
| Co-CGEopenCLIP-based=false2023.05 | 40.8 | 23.3 | 61.2 | 45.8 | — | — | |
| CSPimplementation source=re-implemented2026.03 | 40.6 | 23.3 | — | — | 63.4 | 44.8 | |
| Co-CGECWCLIP-based=false2023.05 | 40.3 | 23.1 | 62 | 44.3 | — | — | |
| SASOW2025.12 | 40.2 | 25.2 | 63.5 | 53.1 | — | — | |
| GIPCOLSetting=Open-World2025.12 | 40.1 | 23.5 | — | — | 65 | 45 | |
| CGECLIP-based=false2023.05 | 39 | 23.1 | 61.7 | 47.7 | — | — | |
| CGE2025.12 | 39 | 23.1 | 61.7 | 47.7 | — | — | |
| CSPSetting=Open-World2025.12 | 38.9 | 22.7 | — | — | 64.1 | 44.1 | |
| CSPCLIP-based=true, Visual Backbone=ViT-L/142023.05 | 38.9 | 22.7 | 64.1 | 44.1 | — | — | |
| VisProd2025.12 | 36.9 | 19.7 | 54.6 | 42.8 | — | — | |
| CompCos2025.12 | 36.9 | 21.3 | 59.3 | 46.8 | — | — | |
| CoOp (a)CLIP-based=true, Visual Backbone=ViT-L/142023.05 | 35.6 | 19.5 | 61.8 | 39.3 | — | — | |
| CompCosCLIP-based=false2023.05 | 34.5 | 18.5 | 53.3 | 44.6 | — | — | |
| SymNet2025.12 | 34.5 | 18.5 | 53.3 | 44.6 | — | — | |
| ProDASetting=Open-World2025.12 | 34.3 | 18.4 | — | — | 63.9 | 34.6 | |
| LE+2025.12 | 30.5 | 16.3 | 60.4 | 36.5 | — | — | |
| AoP2025.12 | 29.4 | 13.7 | 50.9 | 34.2 | — | — | |
| CoOpSetting=Open-World2025.12 | 28.9 | 13.2 | — | — | 52.1 | 31.5 | |
| CoOp (b)CLIP-based=true, Visual Backbone=ViT-L/142023.05 | 28.9 | 13.2 | 52.1 | 31.5 | — | — | |
| CoOP2026.03 | 28.9 | 13.2 | — | — | 52.1 | 31.5 | |
| TMN2025.12 | 21.7 | 8.4 | 55.9 | 18.1 | — | — | |
| CLIPSetting=Open-World2025.12 | 11.2 | 2.2 | — | — | 15.7 | 20.6 | |
| CLIP-ZSCLIP-based=true, Visual Backbone=ViT-L/142023.05 | 11.2 | 2.2 | 15.7 | 20.6 | — | — | |
| CLIP2026.03 | 11.2 | 2.2 | — | — | 15.7 | 20.6 |