Refusal
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Refusal benchmark
97.81Alignment Probability
55
Refusal
0ASR
42
Refusal
86.6Refusal Rate
36
Refusal
69.3Sorry Rate
22
Refusal (test)
64.3BAS (Baseline)
6
REFUSAL
0.567Mean MCC
6
Refusal
0.9MCC
6