ResearchBenchmarksAdversarial Attack Detection on Privacy Extraction Attack DatasetFollow94.5Positive RateICA7.9230.397552.87575.3525Apr 26, 2026Evaluation ResultsMethodMethodLinksPositive RateICADetector=GPT-Safeguard...Detector=GPT-Safeguard, Inference Strategy=cross-instance learning2026.0494.5MEXTRADetector=GPT-Safeguard...Detector=GPT-Safeguard, Inference Strategy=cross-instance learning2026.0493.56SPOREDetector=GPT-Safeguard...Detector=GPT-Safeguard, Inference Strategy=cross-instance learning2026.0411.25