System Prompt Extraction
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
93.5Exact Match Score
20
Jun 4, 2026
91Exact Match Score
20
Jun 4, 2026
0.915Self-Consistency Score
10
Feb 26, 2026
78ASR (TA)
5
Jun 4, 2026
78Attack Success Rate (TA)
5
Jun 4, 2026
77.5Attack Success Rate
4
Jun 4, 2026
75Attack Success Rate
4
Jun 4, 2026
99.7Attack Success Rate (Realistic User)
4
Feb 26, 2026