| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| SALAD Alpaca HarmBench | Residual Paving (oracle) | Target Success99.8 | 18 | 2mo ago | |
| Phi/Qwen judged grid | GoR/DIM ablation | Target Success Rate33.3 | 7 | 2mo ago | |
| N512 objective checks | Residual Paving | Target Success Rate94.8 | 4 | 2mo ago |