Safety Evaluation on SafeBench benchmark leaderboard · SOTA2 Research

Safety Evaluation on SafeBench

99Overall Safety ScoreTELLME NT-Xent

Evaluation Results

Method	Links
TELLME NT-Xent2025.02		99	—	—	—	—	—	—	—	—	—	—
TELLME2025.02		98.1	—	—	—	—	—	—	—	—	—	—
Origin2025.02		97.6	—	—	—	—	—	—	—	—	—	—
SFT2025.02		97.3	—	—	—	—	—	—	—	—	—	—
VROP2026.03		92	98	92	100	96	98	52	98	96	96	98
VROP2026.03		91	100	90	100	98	94	42	96	98	96	100
VROP2026.03		91	94	82	100	100	96	58	96	98	94	90
MM2026.03		89	90	82	95	94	93	53	91	98	96	98
MML2026.03		88	89	84	87	88	83	48	93	91	90	85
MML2026.03		86	92	82	87	90	80	35	89	91	90	88
MML2026.03		86	89	77	85	92	88	52	90	94	88	85
JOOD2026.03		86	85	79	92	82	86	45	89	88	83	90
JOOD2026.03		84	88	77	91	86	82	30	85	88	85	92
JOOD2026.03		83	86	73	90	89	84	47	87	90	84	82
FS2026.03		80	70	70	96	78	84	42	81	98	96	90
FS2026.03		79	74	54	96	72	86	36	70	100	100	96
FS2026.03		72	78	36	80	68	64	28	70	100	100	100
MM2026.03		56	59	48	60	35	43	14	33	84	92	88
MM2026.03		52	47	41	50	32	22	6	30	100	96	98