Multimodal News Classification on NewsMM (test)
91.2AccuracyMultiPress
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MultiPressBackbone=GPT-4o2026.04 | 91.2 | 90.7 | 90.1 | 90.4 | |
| MultiPressBackbone=Qwen2.5-VL-7B2026.04 | 85.3 | 84.7 | 84.1 | 84.4 | |
| Qwen2.5-VL-7BEvaluation Protocol=fine-tuned2026.04 | 80.2 | 79.5 | 79 | 79.2 | |
| GPT-4o-202505132026.04 | 79.6 | 79 | 78.5 | 78.7 | |
| Qwen2.5-VL-7BEvaluation Protocol=zero-shot2026.04 | 77.4 | 76.8 | 76.1 | 76.4 | |
| GLM-4V-9B2026.04 | 70.5 | 69.8 | 69.2 | 69.5 | |
| mPLUG-7B2026.04 | 68.9 | 68.1 | 67.5 | 67.8 | |
| Qwen2.5-VL-3BEvaluation Protocol=fine-tuned2026.04 | 66.8 | 66 | 65.5 | 65.7 | |
| Qwen2.5-VL-3BEvaluation Protocol=zero-shot2026.04 | 62.3 | 61 | 60.5 | 60.7 |