Text Safety Filter Bypass on NSFW-200
31Text Bypass RateOptJail
Evaluation Results
| Method | Links | |
|---|---|---|
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=text-classifier [28]2025.05 | 31 | |
| OptJailT2I Model=SD XL [35], Text Filter=text-classifier [28]2025.05 | 31 | |
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=DeepSeek-V3 [9]2025.05 | 83.5 | |
| OptJailT2I Model=SD XL [35], Text Filter=DeepSeek-V3 [9]2025.05 | 83.5 | |
| OptJailT2I Model=DALL·E 3 [3], Text Filter=unknown2025.05 | 84 | |
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=GPT-4.1 [34]2025.05 | 97 | |
| OptJailT2I Model=SD XL [35], Text Filter=GPT-4.1 [34]2025.05 | 97 | |
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=Detoxify [20]2025.05 | 97.5 | |
| OptJailT2I Model=SD XL [35], Text Filter=Detoxify [20]2025.05 | 97.5 | |
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=text-match [18]2025.05 | 99 | |
| OptJailT2I Model=FLUX.1-schnell [4], Text Filter=ShieldLM-7B-internlm2 [50]2025.05 | 99 | |
| OptJailT2I Model=SD XL [35], Text Filter=text-match [18]2025.05 | 99 | |
| OptJailT2I Model=SD XL [35], Text Filter=ShieldLM-7B-internlm2 [50]2025.05 | 99 |