Harmful Content Detection on HoliSafe-Bench
75.6AUPRCQwen3-VL 8B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-VL 8BPrompt Selection Strategy=Max152026.05 | 75.6 | |
| Qwen3-VL 4BPrompt Selection Strategy=Max152026.05 | 75.2 | |
| Qwen3-VL 4BPrompt Selection Strategy=Sel2026.05 | 74.6 | |
| Qwen3-VL 4BPrompt Selection Strategy=Rand2026.05 | 74.5 | |
| InternVLPrompt Selection Strategy=Max152026.05 | 73.4 | |
| Qwen3-VL 8BPrompt Selection Strategy=Rand2026.05 | 72.6 | |
| Qwen3-VL 4BPrompt Selection Strategy=MeanEns2026.05 | 71.8 | |
| Qwen3-VL 8BPrompt Selection Strategy=MeanEns2026.05 | 71.5 | |
| Qwen3-VL 8BPrompt Selection Strategy=Med152026.05 | 70.6 | |
| GemmaPrompt Selection Strategy=Max152026.05 | 70.3 | |
| Qwen3-VL 4BPrompt Selection Strategy=Med152026.05 | 70.1 | |
| InternVLPrompt Selection Strategy=Rand2026.05 | 69.6 | |
| InternVLPrompt Selection Strategy=MeanEns2026.05 | 68.7 | |
| Qwen3-VL 4BPrompt Selection Strategy=Mean152026.05 | 68.5 | |
| GemmaPrompt Selection Strategy=Rand2026.05 | 67.7 | |
| Qwen3-VL 8BPrompt Selection Strategy=Mean152026.05 | 67.4 | |
| MiniCPMPrompt Selection Strategy=Max152026.05 | 67.2 | |
| MiniCPMPrompt Selection Strategy=MeanEns2026.05 | 65.5 | |
| GemmaPrompt Selection Strategy=Med152026.05 | 65.4 | |
| InternVLPrompt Selection Strategy=Med152026.05 | 64.2 | |
| MiniCPMPrompt Selection Strategy=Rand2026.05 | 64.1 | |
| InternVLPrompt Selection Strategy=Sel2026.05 | 63.4 | |
| InternVLPrompt Selection Strategy=Mean152026.05 | 63.4 | |
| Qwen30BPrompt Selection Strategy=Max152026.05 | 63 | |
| GemmaPrompt Selection Strategy=Mean152026.05 | 62.6 | |
| LlamaPrompt Selection Strategy=Max152026.05 | 61.4 | |
| Qwen30BPrompt Selection Strategy=Rand2026.05 | 59.9 | |
| MiniCPMPrompt Selection Strategy=Med152026.05 | 59.7 | |
| GemmaPrompt Selection Strategy=MeanEns2026.05 | 59 | |
| MiniCPMPrompt Selection Strategy=Mean152026.05 | 58.5 | |
| Qwen30BPrompt Selection Strategy=MeanEns2026.05 | 57.6 | |
| Qwen30BPrompt Selection Strategy=Mean152026.05 | 56 | |
| MiniCPMPrompt Selection Strategy=Sel2026.05 | 55.6 | |
| Qwen30BPrompt Selection Strategy=Med152026.05 | 55.5 | |
| LlamaPrompt Selection Strategy=MeanEns2026.05 | 53.9 | |
| LlamaPrompt Selection Strategy=Rand2026.05 | 53.4 | |
| LlamaPrompt Selection Strategy=Med152026.05 | 52.9 | |
| Qwen30BPrompt Selection Strategy=Sel2026.05 | 51.7 | |
| LlamaPrompt Selection Strategy=Mean152026.05 | 51.2 | |
| Qwen30BPrompt Selection Strategy=Min152026.05 | 49.2 | |
| LlamaPrompt Selection Strategy=Sel2026.05 | 44 | |
| Qwen3-VL 4BPrompt Selection Strategy=Min152026.05 | 43.4 | |
| Qwen3-VL 8BPrompt Selection Strategy=Sel2026.05 | 41.3 | |
| Qwen3-VL 8BPrompt Selection Strategy=Min152026.05 | 41.3 | |
| MiniCPMPrompt Selection Strategy=Min152026.05 | 40.9 | |
| LlamaPrompt Selection Strategy=Min152026.05 | 40.2 | |
| InternVLPrompt Selection Strategy=Min152026.05 | 36.3 | |
| GemmaPrompt Selection Strategy=Sel2026.05 | 33.2 | |
| GemmaPrompt Selection Strategy=Min152026.05 | 33.2 |