List your company

Jailbreak Robustness (BEAST) on HarmBench

9.7Median ASRSAE-augmented models
9.31411.919514.52517.1305Apr 20, 2026

Evaluation results

MethodLinks
SAE-augmented modelsModel Configuration=SAE2026.049.70.0411
BaselineModel Configuration=BASE2026.0419.35—