Prompt continuation on Real Toxicity Prompts Profanity
46.8RSRBRT (s+r)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BRT (s+r)Victim Model=GPT-3, Query Limit=10,0002023.05 | 46.8 | 19.6 | |
| BRT (s)Victim Model=GPT-3, Query Limit=10,0002023.05 | 40.4 | 19.6 | |
| RandVictim Model=GPT-3, Query Limit=10,0002023.05 | 24.1 | 22.1 | |
| Offensive Top-NQVictim Model=GPT-3, Query Limit=10,0002023.05 | 19.4 | 24.5 |