Red Teaming

Benchmarks

Dataset Name	SOTA Method	Metric
HarmBench	AutoDAN-Turbo	ASR96.3	244	1mo ago
Violence prompts	UnlearnDiffAtk	Failure Rate (FR)0	48	21d ago
I2P Nudity prompts	Ring-A-Bell	Failure Rate (FR)0	48	21d ago
50 harmful goals (Manual evaluation)	PAIR	Hard ASR100	30	3mo ago
CatQA	SafeTransformer	ASR0	20	2mo ago
AdversarialQA	SafeTransformer	ASR0	20	2mo ago
Religious Discrimination principle v1 (test)	QCI	Mean Best Category Score5.32	12	3mo ago
Illegal Activity principle v1 (test)		Mean Score (Best Category)-2.73	12	3mo ago
AI Supremacy principle v1 (test)	CRL	Mean Best Category Score11.7	12	3mo ago
AdvBench (test)	AMIS	ASR88	8	2mo ago
DailyDialog against DialoGPT-large	BRT (e+r)	RSR40	8	3mo ago
DailyDialog against BB-3B	BRT (e+r)	RSR40.2	8	3mo ago
ConvAI2 (filtered hard positive)	BRT (e+r)	RSR2,120	7	3mo ago
Bloom ZS (filtered hard positive)	BRT (e+r)	RSR15.6	7	3mo ago
BAD Against Friend Chat (test)	BRT (e)	RSR64.2	7	3mo ago
BAD Against Marv (test)	BRT (s+r)	RSR88.1	7	3mo ago
GPT-OSS 20B	PAIR	Coverage63.2	5	26d ago
Llama-3-8B	Ours (ME)	Coverage63.04	5	26d ago
Web-Augmented LLM Red-Teaming Evaluation Set	CREST-Search	Detection Rate80.5	5	1mo ago
Korean red teaming dataset (test)	Exaone-3.5-2.4B-inst	Attack Success Rate0.5797	5	3mo ago
HarmBench Claude-Sonnet-3.5 (held-out test)	AGENTICRED	ASR60	5	3mo ago
HarmBench Llama-3-8B (test)	AGENTICRED	ASR0.98	5	3mo ago
HarmBench Llama-2-7B (test)	AutoDAN-Turbo	ASR36	5	3mo ago
GPT-5 Mini	Ours (ME)	Coverage72.32	4	26d ago
KT RAIC proprietary Korean red-teaming dataset		Attack Success Rate54	4	2mo ago

Showing 25 of 35 rows