Deepfake Detection on Generalization Evaluation Suite

100ProGAN AccuracyGragn. et al.

Evaluation results

Method
Gragn. et al.2025.12	100	93.27	91.75	97.55	94.13	99.65	97.25	89.75	97.47	67.45	60.65	69.38	67.3	62.33	59.7	57.75	65.31	50.02	76.59
Corvi et al.2025.12	100	95.85	90.35	98.4	92.46	99	97.65	84.9	82.79	65.3	69.3	58.98	53.1	58.83	55.7	52.1	59.38	50.11	72.72
Wang et al.2025.12	99.9	67.65	79.5	72.65	76.63	89.72	82.1	77.05	80.68	56.45	55.05	61.15	62.9	54.03	52.5	53.4	52.67	49.68	66.09
Khan et al.2025.12	99.88	94.75	97.45	95.3	95.47	99.12	93.35	78.35	93.11	94.55	92	94.27	81.65	89.18	67.7	71.6	77.11	70.16	88.72
Khan et al.2025.12	99.83	93.8	95.6	93.5	93.43	99.15	95.25	82.95	93.11	94.95	91.5	92.88	84.3	88.16	76.45	77.8	78.45	74.66	89.45
Wang et al.2025.12	99.65	58.13	77.8	50.3	75.56	79.99	69.8	62.3	53.42	51.05	51.9	54.33	52.35	51.35	50.15	51	51.46	50.02	59.18
Khan et al.2025.12	99.6	77.38	71.55	98.4	65.7	100	94.85	95.3	99.89	94.4	93.2	88.78	92.35	95.17	91.75	97.35	76.46	52.11	88.74
Ojha et al.2025.12	98.94	94.48	94.2	57.75	94.65	87.49	85.55	83.4	75.42	89.45	89.2	82.15	79	87.8	81.9	74.15	62.71	64.3	82.84
Khan et al.2025.12	98.5	91.75	91	98.2	88.08	94.42	81.4	71.7	94.11	91.05	85.8	90.55	79.05	87.42	77.3	83.85	69.37	68.3	86.26
AdaptPrompt_v22025.12	98.46	70.47	70.1	98.1	69.99	97.72	94.35	89.8	97.26	97	95.75	90.7	93.5	92.33	98.1	97.9	57.51	58.18	88.93
AdaptPrompt_v02025.12	98.4	69.72	71.35	96.2	66.65	93.87	95.35	91	97.84	94.55	94.65	88.18	89.2	91.75	96.2	96	53.63	54.95	87.34
Adapter_v02025.12	98.24	73.95	79.4	98.35	74.52	95.42	95.2	86.75	98.89	96.85	96.45	88.68	91	92.18	98.35	97.85	54.45	57.11	88.88
Prompt Tuning_v22025.12	98.2	70.15	72.9	96.15	64.22	97.85	95.2	90.4	99.85	94.25	93.65	87.05	88.9	90.27	96.15	95.7	52.3	54.12	87.11
Adapter_v22025.12	97.45	80.67	80.9	99.1	79.98	97.72	90.4	79.45	99.05	95.85	93.5	81.63	85.75	87.73	99.15	98.25	55	54.82	87.31
Prompt Tuning_v02025.12	93.76	73.65	73.55	97.6	72.09	90.4	95.05	86.3	98.84	95.5	95.65	84.98	89.85	90.62	97.5	97.1	51.58	52	86.96
Linear Probing_v22025.12	93.24	68.58	65.5	93.95	64.83	87.97	93.7	86.7	96.37	94.8	97.7	72.88	89.05	83.52	95.9	95.8	52.43	51.21	82
Fine Tuning_v12025.12	92.51	61.23	61.1	94.55	51.83	96.17	88.75	90.4	98.05	93.85	93.05	85.23	90.6	94.28	94.55	94.55	50.19	51.32	84.98
Fine Tuning_v22025.12	92.11	64.78	65.55	98.1	53.57	94.2	87.7	87.75	99.26	97.1	96.75	88.23	92.95	97.32	98.1	97.8	50.73	52	87.08
Linear Probing_v02025.12	89.14	64.95	66.85	95.1	60.13	79.83	95.15	86.7	97.58	94.15	92.65	71.57	89.45	84.07	95.25	95.25	49.02	51.04	80.59
Fine Tuning_v02025.12	88.79	65.12	63.5	98.5	53.78	96.82	87.25	89.8	99.68	97.65	97.7	87.55	94.5	98.05	98.4	98.35	52.43	51.21	87.39
Corvi et al.2025.12	50.94	51.82	46.2	49.25	50.86	48.02	59.4	50.95	50.05	77.65	87	59.83	50.95	99.25	99.25	93.1	69.87	48.14	66.4