LLM Skill-Security corpus
Benchmarks
TaskDatasetSOTA resultTrendResults
LLM Skill-Security corpus 10 synthetic scenarios
97.14Empirical Accuracy
4
LLM Skill-Security corpus 93 malicious samples
97.6Empirical Accuracy
4
LLM Skill-Security corpus 250 tampered artifacts
95.1Empirical Accuracy
2
LLM Skill-Security corpus 250 tampered artifacts
100Mechanism-level Guarantee
1
LLM Skill-Security corpus discrete-event voting simulation
40Highest Tolerable Ratio of Malicious Auditors
1