Face Anti-Spoofing on Unified Face Attack Detection Protocol 3 (test) benchmark leaderboard · SOTA2 Research

Face Anti-Spoofing on Unified Face Attack Detection Protocol 3 (test)

100Accuracy (Real Face)GLM-4.6V

Evaluation Results

Method	Links
GLM-4.6V2026.05		100	0	15	85	1.67	93.3	0	100	0	100	0	100	0	100	8.4	91.3
UniShield2026.05		99.3	0.67	99.7	0.33	99.3	0.33	98.4	1	97.3	1.33	100	0	100	0	99.4	0.38
UniShield2026.05		99	1	98.8	0.83	94	3	97.3	1.34	96.7	1.67	99.3	0.33	99.9	0.06	98.6	0.79
HiPTune2026.05		97.3	2.67	99.8	0.17	100	0	88.1	11.9	98	1	99	0.5	99.9	0.06	97.3	2.58
Gemini-3-Pro2026.05		95	5	60	20	66.7	18.3	19	40.5	0	100	13.3	43.4	17.3	41.4	29.5	39
UniAttackDetection2026.05		85.3	7.33	99.3	0.33	98.7	0.67	98.9	0.56	98.7	0.67	98.2	0.92	100	0	98.4	0.82
CoOp2026.05		78	11	99.3	0.33	99	0.5	96.1	1.96	99.7	0.17	93.8	3.08	100	0	96.9	1.55
GPT-5.22026.05		61.7	38.3	38.3	45	2	53.3	0	100	0	100	0	100	0.29	49.9	9.05	68
CLIP2026.05		25.7	37.2	99.3	0.33	99.3	0.33	98.9	0.56	98.7	0.67	99	0.5	100	0	94.7	2.67
CDCN2026.05		1.33	49.3	76.2	11.9	66.3	16.8	38.5	30.7	25.3	37.3	16.5	41.8	3.28	48.4	26.3	36.9