Image-to-Text Retrieval on MS-COCO (Spearman S, R2)
0.981Spearman SGeoFlowVLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GeoFlowVLMPre-training Dataset=CC1M, Evaluation Protocol=Zero-shot2026.05 | 0.981 | 0.882 | |
| ProbVLMPre-training Dataset=CC1M, Evaluation Protocol=Zero-shot2026.05 | 0.717 | 0.621 | |
| AsymVLM-PSDPre-training Dataset=CC1M, Evaluation Protocol=Zero-shot2026.05 | 0.35 | 0.123 | |
| AsymVLM-vMFPre-training Dataset=CC1M, Evaluation Protocol=Zero-shot2026.05 | 0.35 | 0.123 | |
| GroVEPre-training Dataset=CC1M, Evaluation Protocol=Zero-shot2026.05 | 0.105 | 0.025 |