List your company

Jailbreak Detection on Reconstructed Embedding Space Jailbreak (100 sampled sentences per label)

100AccuracyLabel Forensics
83.3687.689296.32Dec 1, 2025

Evaluation results

MethodLinks
Label ForensicsLabel=0 (Clean Prompt)2025.12100
Label ForensicsLabel=Avg.2025.1292
Label ForensicsLabel=1 (Jailbreak)2025.1284