Multimodal Retrieval on MAG Toys benchmark
87.4R@1GOMA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GOMAMethod Category=Our Method, Feature Backbone=Qwen2-VL2026.05 | 87.4 | 99.5 | 99.9 | |
| DGFMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 85.8 | 98.7 | 99.6 | |
| DMGCMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 85.2 | 98.5 | 99 | |
| SmartCLIPMethod Category=Pairwise Image-Text Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 56.7 | 83.1 | 89.4 | |
| VSE++Method Category=Pairwise Image-Text Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 53.6 | 81 | 87.8 | |
| CLIP-RefineMethod Category=Pairwise Image-Text Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 53.6 | 80.6 | 86.9 | |
| D2S-VSEMethod Category=Pairwise Image-Text Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 50 | 78.3 | 86.3 | |
| LinearMethod Category=Direct Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 38.8 | 66.8 | 76.8 | |
| LGMRecMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 34.4 | 69.6 | 81.4 | |
| MMGCNMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 18.8 | 50.6 | 67.1 | |
| MGATMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 15.9 | 43.8 | 59.9 | |
| GSMNMethod Category=Multimodal Graph Retrieval & Recent MAG Models, Feature Backbone=Qwen2-VL2026.05 | 13.5 | 38.6 | 53.1 | |
| RawMethod Category=Direct Retrieval Baselines, Feature Backbone=Qwen2-VL2026.05 | 12.1 | 26 | 33.4 |