Binary Classification on PHTV (test)
94.29AccuracyQwen3-VL-8B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-VL-8BLearning Paradigm=Fine-tuned Models, Tuning Technique=LoRA2026.05 | 94.29 | 96.41 | 92 | 94.15 | 3.43 | |
| Early Fusion (Text+Video)Modality=Text+Video, Learning Paradigm=Fine-tuned Pre-trained Models2026.05 | 92.57 | 94.08 | 90.86 | 92.44 | 5.71 | |
| Qwen-VL-MAXLearning Paradigm=In-Context Learning (ICL), Shot-count=128-shot2026.05 | 79.66 | 82.1 | 76 | 78.93 | 16.67 | |
| BERT (Text-only)Modality=Text-only, Learning Paradigm=Fine-tuned Pre-trained Models2026.05 | 78.57 | 76.88 | 81.71 | 79.22 | 24.57 | |
| Late Fusion (Text+Video)Modality=Text+Video, Learning Paradigm=Fine-tuned Pre-trained Models2026.05 | 77.43 | 86.92 | 64.57 | 74.1 | 9.71 | |
| VideoMAE (Video-only)Modality=Video-only, Learning Paradigm=Fine-tuned Pre-trained Models2026.05 | 72.29 | 71.91 | 73.14 | 72.52 | 28.57 | |
| Claude-Opus-4.1Learning Paradigm=In-Context Learning (ICL), Prompting Strategy=CoT-EN2026.05 | 63.87 | 89.83 | 30.81 | 45.89 | 3.45 | |
| GPT-4-TurboLearning Paradigm=In-Context Learning (ICL), Prompting Strategy=CoT-EN2026.05 | 54.29 | 66.67 | 17.14 | 27.27 | 8.57 | |
| Gemma3-12BLearning Paradigm=In-Context Learning (ICL)2026.05 | 53.43 | 87.5 | 8 | 14.66 | 1.14 |