Red-teaming on AdvBench current (full)
87.3ASR (%)OTTER-MLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OTTER-MLMModel=GPT-3.5-turbo2026.06 | 87.3 | 84.2 | |
| OTTER-RVModel=avg.2026.06 | 84 | — | |
| GPT-4o paraphraseModel=avg.2026.06 | 77 | — | |
| OTTER-MLMModel=Avg.2026.06 | 75.6 | — | |
| OTTER-MLMModel=GPT-4o-mini2026.06 | 74.2 | 70.2 | |
| OTTER-MLMModel=GPT-4-turbo2026.06 | 71.3 | 67 | |
| OTTER-MLMModel=GPT-4o2026.06 | 69.8 | 65.9 | |
| Direct promptModel=GPT-3.5-turbo2026.06 | 13.1 | 10.3 | |
| Direct promptModel=GPT-4o-mini2026.06 | 7.4 | 5 | |
| Direct promptModel=Avg.2026.06 | 7 | — | |
| Direct promptModel=GPT-4o2026.06 | 6.6 | 4.4 | |
| Direct promptModel=GPT-4-turbo2026.06 | 0.7 | 0 |