| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| HumanEval | Perplexity Filter | False Positive Rate0 | 32 | 3mo ago | |
| OR-Bench | Perplexity Filter | False Positive Rate0 | 26 | 3mo ago | |
| ULSPB 350 interaction runs | StateGuard (Base-Ensemble) | HS Rate0.02 | 24 | 2mo ago | |
| AlpacaEval | Llama Guard | False Positive Rate0 | 24 | 3mo ago | |
| GSM8K | Perplexity Filter | False Positive Rate0 | 24 | 3mo ago | |
| SafeEditBench | PolicyShiftGuard-7B | Macro F161.7 | 9 | 18d ago | |
| UnSafeBench | PolicyShiftGuard-7B | F1 Score64.1 | 9 | 18d ago |
