Harmful Requests
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Harmful Requests RepE Attack
91.2Compliance Rate
6
Harmful Requests Input Embed
92.1Compliance Rate
6
Harmful Requests Prefilling
95Compliance Rate
6
Harmful Requests Multilingual
34.1Compliance Rate
6
Harmful Requests GCG
0.887Compliance Rate
6
Harmful Requests PAIR
69.5Compliance Rate
6
Harmful Requests TAP-T
85.8Compliance Rate
6
Harmful Requests AutoDAN
93.4Compliance Rate
6
Harmful Requests Manual Attack
77.4Compliance Rate
6
Harmful Requests No Attack
57.8Compliance Rate
6