Deepfake Detection on FF++ LOO benchmark leaderboard · SOTA2 Research

Deepfake Detection on FF++ LOO

67.44Acc @ 0.5DAG-FDD

Evaluation Results

Method	Links
DAG-FDD2026.06		67.44	50.81	17.78	4.48	0.2226	50	12
DAG-FDD + FFT2026.06		63.76	55.46	42.22	46.81	0.8903	42	11
DAID + FFT2026.06		63.62	52.16	35.56	48.46	0.8402	28	11
DAW-FDD + FFT2026.06		63.19	51.51	28.89	44.83	0.7372	40	9
PG-FDD + FFT2026.06		61.75	49.86	50	42.72	0.9272	30	10
FFT2026.06		60.22	54.88	40.48	46.12	0.866	36	11
BPFA2026.06		49.57	52.31	60	38.19	0.9819	30.56	12
BASE2026.06		47.8	52.4	65	42.36	1.0736	27.78	—
Fair-FLIP2026.06		46.85	52.57	65	36.81	1.0181	27.78	4
FF Discover2026.06		45.08	—	57.69	39.42	0.9711	40.22	11
FF Max2026.06		43.84	—	85.29	73.66	1.5895	30.56	10
DAID2026.06		42.05	52.25	76.19	47.6	1.2379	27.87	4
PG-FDD2026.06		39.61	56.44	32.69	37.22	0.6991	31.25	4
DAW-FDD2026.06		34.96	49.55	34.62	34.04	0.6866	30	3