Multi-modal Retrieval-Augmented Generation on MC-SEARCH Image-Initiated Chain 1.0 (test)
47.61F1Gemini-2.5-Pro
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Gemini-2.5-ProAccess=CLOSE, Enhanced=None2026.03 | 47.61 | 42.76 | 3.18 | 25.9 | 1.05 | 69.83 | |
| Qwen2.5-VL-7B + SEARCH-ALIGNAccess=OPEN, Enhanced=SEARCH-ALIGN2026.03 | 45.7 | 28.05 | 2.23 | 33.59 | 0.7 | 60.95 | |
| Gemini-2.5-FlashAccess=CLOSE, Enhanced=None2026.03 | 44.1 | 37.38 | 3.01 | 31.46 | 2.91 | 72.39 | |
| InternVL3.5-8B + SEARCH-ALIGNAccess=OPEN, Enhanced=SEARCH-ALIGN2026.03 | 42.27 | 32.65 | 2.53 | 32.49 | 0.94 | 63.86 | |
| InternVL3.5-8BAccess=OPEN, Enhanced=None2026.03 | 39.11 | 29.49 | 2.27 | 22.59 | 1.58 | 63.86 | |
| Claude-3.7-SonnetAccess=CLOSE, Enhanced=None2026.03 | 37.8 | 33.09 | 2.6 | 27.31 | 1.18 | 72.62 | |
| GPT-4o-MiniAccess=CLOSE, Enhanced=None2026.03 | 36.49 | 34.18 | 2.63 | 27.51 | 1.46 | 68.29 | |
| Qwen2.5-VL-7BAccess=OPEN, Enhanced=None2026.03 | 26.3 | 8.65 | 1.34 | 16.51 | 4.04 | 60.95 |