ResearchBenchmarksJailbreak Attack on Harmful Prompts model-averagedJailbreak Attack on Harmful Prompts model-averagedFollow4.8Model Averaged ASR (0.8)AmpleGCG+2.27219.33636.453.464Jun 2, 2026Evaluation resultsMethodMethodLinksModel Averaged ASR (0.8)Scaled ScoreAmpleGCG+2026.064.80PAIR2026.0616.60.56BON2026.0623.20.87GCG+2026.0630.31.21PAIR+2026.0636.31.5INPAINTING2026.0643.31.83JR1+2026.0651.62.22IHO2026.06683