Diversity Assessment on Red Teaming Evaluation B=1000
0.0007Kruskal-Wallis p-valueToxSearch vs ToxSearch-S
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ToxSearch vs ToxSearch-SOutcome=Semantic spread (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0007 | 0.128 | 0.128 | 0.51 | |
| ToxSearch vs RainbowPlusOutcome=Semantic spread (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0007 | 5.83 | 0.0018 | 1 | |
| ToxSearch-S vs RainbowPlusOutcome=Semantic spread (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0007 | 5.83 | 0.0018 | 1 | |
| ToxSearch vs ToxSearch-SOutcome=DBSCAN clusters (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0261 | 0.0695 | 0.139 | 0.571 | |
| ToxSearch vs RainbowPlusOutcome=DBSCAN clusters (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0261 | 0.433 | 0.433 | 0.245 | |
| ToxSearch-S vs RainbowPlusOutcome=DBSCAN clusters (top-K), B (Budget)=1000, n (runs per method)=72026.06 | 0.0261 | 0.0153 | 0.0458 | 0.776 |