Safety Evaluation on Harmful Prompts Text-only baseline
0Text ASRclaude-sonnet-4.5-20250929
Evaluation Results
| Method | Links | |
|---|---|---|
| claude-sonnet-4.5-20250929Model Type=Proprietary2026.05 | 0 | |
| claude-haiku-4.5-20251001Model Type=Proprietary2026.05 | 8 | |
| gpt-4.1-2025-04-14Model Type=Proprietary2026.05 | 12.7 | |
| doubao-seed-1.6-251015Model Type=Proprietary2026.05 | 17.3 | |
| glm-4.5vModel Type=Open Source2026.05 | 29.2 | |
| qwen3-vl-235b-a22b-thinkingModel Type=Open Source2026.05 | 31.4 | |
| qwen3-vl-plusModel Type=Proprietary2026.05 | 35.3 | |
| qwen3-vl-32b-thinkingModel Type=Open Source2026.05 | 36.7 | |
| gemini-2.5-flashModel Type=Proprietary2026.05 | 47.5 | |
| intern-s1Model Type=Open Source2026.05 | 90.8 |