Hateful video detection on Ex-HateMM (test)
90.37AccuracyQwen2.5-VL (7B) w/ IARE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-VL (7B) w/ IAREModality=LLM-based, Enhancement=Implicit reasoning enhancement2026.06 | 90.37 | 85.42 | 92.13 | 90.14 | |
| GLM4.1V (9B) w/ IAREModality=LLM-based, Enhancement=Implicit reasoning enhancement2026.06 | 87.61 | 84.88 | 83.91 | 87.06 | |
| Qwen2.5-VL (7B) w/ SFTModality=LLM-based, Fine-tuning=Supervised Fine-Tuning2026.06 | 86.7 | 87.18 | 78.16 | 85.86 | |
| TCLModality=Multimodal2026.06 | 85.32 | 78.95 | 86.21 | 84.91 | |
| GLM4.1V (9B) w/ SFTModality=LLM-based, Fine-tuning=Supervised Fine-Tuning2026.06 | 84.4 | 82.72 | 77.01 | 83.54 | |
| MoREModality=Multimodal2026.06 | 83.41 | 81.78 | 82.35 | 82.35 | |
| TTMModality=Multimodal2026.06 | 79.8 | 74.2 | 75.8 | 79 | |
| Qwen2.5-VL (7B)Modality=LLM-based, Fine-tuning=Zero-shot2026.06 | 77.13 | 65.74 | 88.32 | 77.01 | |
| ViTModality=Unimodal2026.06 | 74.8 | 69.5 | 65.6 | 67.2 | |
| GLM4.1V (9B)Modality=LLM-based, Fine-tuning=Zero-shot2026.06 | 74.31 | 83.67 | 46.07 | 70.31 | |
| BERTModality=Unimodal2026.06 | 73.5 | 67.5 | 66.7 | 72.2 | |
| CSIDModality=Multimodal2026.06 | 73.2 | 72 | 72.3 | 71.4 | |
| MFCCModality=Unimodal2026.06 | 67.5 | 59.3 | 67.9 | 66.5 | |
| MulTModality=Multimodal2026.06 | 65.71 | 43.18 | 65.71 | 52.12 |