Red Teaming on Bloom ZS (filtered hard positive)
15.6RSRBRT (e+r)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BRT (e+r)Target Model=BB-3B, Query limit (NQ)=200002023.05 | 15.6 | 0.457 | |
| BRT (s+r)Target Model=BB-3B, Query limit (NQ)=200002023.05 | 6.4 | 0.501 | |
| SLTarget Model=BB-3B, Query limit (NQ)=20000, Backbone=OPT-1.3B2023.05 | 5.4 | 0.604 | |
| SFSTarget Model=BB-3B, Query limit (NQ)=20000, Backbone=OPT-1.3B2023.05 | 3.3 | 0.514 | |
| Offensive Top-NQTarget Model=BB-3B, Query limit (NQ)=200002023.05 | 3.1 | 0.502 | |
| SFSTarget Model=BB-3B, Query limit (NQ)=20000, Backbone=Bloom2023.05 | 2.6 | 0.523 | |
| RandTarget Model=BB-3B, Query limit (NQ)=200002023.05 | 0.6 | 0.519 |