StrongREJECT
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
StrongREJECT Llama 3.3 70B
39.8Bin-ASR
5
StrongREJECT Claude Opus 4.1
13.9Bin-ASR
5
StrongREJECT Deepseek-R1
94.5Bin-ASR
5
StrongREJECT OpenAI o1
82.5Bin-ASR
5
StrongREJECT OpenAI o3
23Bin-ASR
5
StrongReject
99.79Robustness Score
5
StrongREJECT (test)
100Overall Score
5
strongREJECT
99.4Acceptance Rate
4
StrongReject
83Goodness@0.1
4
StrongReject
99.7Jailbreak Resistance: Illicit Crime
4
StrongReject (test)
100StrongReject Score
4
StrongREJECT AutoDAN-Turbo attack (test)
87LlamaGuard Score
3
StrongREJECT (sample of 40 prompts)
26.87Mean StrongREJECT Score
3
StrongReject
95.8F1 Score
2