Single-conditional retrieval on CLAY-Object
64.9Color Success RateCLAY
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLAYBackbone=SigLIP-L2026.04 | 64.9 | 94 | 83.3 | |
| CLAYBase VLM=SigLIP-B2026.04 | 63.3 | 94.3 | 81.9 | |
| CLAYBase VLM=CLIP-B2026.04 | 47.8 | 93.4 | 65 | |
| CLAYBackbone=CLIP-L2026.04 | 47.2 | 95.1 | 78.2 | |
| MagicLens2026.04 | 17.3 | 59.9 | 39.3 | |
| SigLIP-LBackbone=SigLIP-L2026.04 | 15.3 | 67.5 | 60.8 | |
| SigLIP-BBase VLM=SigLIP-B2026.04 | 14.7 | 69.6 | 62 | |
| InstructBLIP2026.04 | 14.4 | 75.6 | 61.4 | |
| CLIP-LBackbone=CLIP-L2026.04 | 12.9 | 70.3 | 48.8 | |
| CLIP-BBase VLM=CLIP-B2026.04 | 12.9 | 68.9 | 42.4 | |
| GeneCIS2026.04 | 12.5 | 80.1 | 38.4 | |
| SEARLE2026.04 | 12.1 | 69.1 | 38.8 |