Multi-image Retrieval on MS-COCO Random
94.92AccuracyLAD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LADBackbone=LLaVA-OneVision-8B2026.05 | 94.92 | 3.33 | 84.1 | |
| LADBackbone=Qwen2.5-VL-3B2026.05 | 93.1 | 3.68 | 76.8 | |
| LADBackbone=InternVL-32026.05 | 90.54 | 13.89 | 66.7 | |
| PriDeBackbone=InternVL-32026.05 | 54.52 | 29.87 | 4.3 | |
| VanillaBackbone=InternVL-32026.05 | 51 | 31.27 | 5.8 | |
| InstructionBackbone=InternVL-32026.05 | 45.44 | 29.28 | 3.1 | |
| PriDeBackbone=LLaVA-OneVision-8B2026.05 | 41.52 | 34.51 | 1 | |
| PriDeBackbone=Qwen2.5-VL-3B2026.05 | 34.62 | 30.4 | 2.3 | |
| CoTBackbone=Qwen2.5-VL-3B2026.05 | 31.94 | 29.39 | 2.91 | |
| CoTBackbone=InternVL-32026.05 | 31.84 | 12.5 | 0.7 | |
| VanillaBackbone=LLaVA-OneVision-8B2026.05 | 31.56 | 38.82 | 0.1 | |
| CoTBackbone=LLaVA-OneVision-8B2026.05 | 30.5 | 37.19 | 1.3 | |
| InstructionBackbone=Qwen2.5-VL-3B2026.05 | 28.86 | 28.24 | 1.3 | |
| VanillaBackbone=Qwen2.5-VL-3B2026.05 | 28.68 | 28.86 | 1.2 | |
| InstructionBackbone=LLaVA-OneVision-8B2026.05 | 26.72 | 35.49 | 0.2 |