Jailbreak Prompt Quality Evaluation on 500 randomly sampled prompts benchmark leaderboard · SOTA2 Research

Jailbreak Prompt Quality Evaluation on 500 randomly sampled prompts

81SimilarityPAIR

Evaluation Results

Method	Links
PAIR2026.05		81	43.66	—	31	—	7	74	74
GCG2026.05		80	32.98	25	30	40	10	92	54
GCG2026.05		65	21.22	24	33	43	13	98	78
PAIR2026.05		64	39.35	—	—	40	8	66	56
SAA2026.05		50	38.4	26	32	41	19	100	100
SAA2026.05		49	29.23	23	32	—	15	100	100
Puzzler2026.05		40	33.37	25	33	41	15	100	100
Puzzler2026.05		40	33.37	25	33	41	15	100	100
DrAttack2026.05		36	17.53	24	33	40	16	88	96
DrAttack2026.05		33	19.13	24	33	41	14	70	54
IJP2026.05		16	20.87	22	—	—	—	100	100
ZULU2026.05		16	1,052.22	28	34	44	4	100	96
ZULU2026.05		14	1,155.85	—	—	44	4	100	98
IJP2026.05		12	53.65	—	—	—	3	100	100
Base642026.05		11	87.43	—	—	—	3	100	100
Base642026.05		11	89.54	—	—	—	3	100	100