Multi-class Classification on MAMI Task B
79.6Macro-F1KID
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| KID2026.01 | 79.6 | 87.5 | |
| Qwen2.5-VL-7BArchitecture=Dual-Head2026.01 | 75.36 | 83.56 | |
| Gemini-2.5-FlashProtocol=Few-shot2026.01 | 75.34 | 74.22 | |
| Qwen2.5-VL-7BArchitecture=Dual-Head, Input=Caption2026.01 | 74.54 | 82.86 | |
| Gemini-2.5-FlashProtocol=Zero-shot2026.01 | 73.44 | 73.7 | |
| Previous SOTA2026.01 | 73 | — | |
| Qwen2.5-VL-7BProtocol=SFT2026.01 | 70.35 | 78.32 |