AI-Generated Image Detection on CNNSpot-Generalization-Suite trained on ProGAN (test) benchmark leaderboard · SOTA2 Research

AI-Generated Image Detection on CNNSpot-Generalization-Suite trained on ProGAN (test)

99.99Pro-GAN Detection AccuracyCNNSpot

Evaluation Results

Method	Links
CNNSpot2025.12		99.99	87.59	71.18	89.95	94.6	81.44	51.69	50	50.82	50.88	60.2	57.85	50.77	51.13	56.2	66.95
VIB-Net2025.12		99.99	99	95.75	91.25	98.95	99.7	83.2	70.5	71.55	70	71.45	69.4	61.25	75.9	86.65	82.97
Fusing2025.12		99.9	87.01	77.32	85.21	97.04	76.95	53.76	54.56	51.04	51.34	56.5	57.15	52.12	51.67	55.09	67.11
Univfd2025.12		99.9	98.5	94.5	84.4	95.85	99.5	67.4	56.5	63.1	63.57	66.9	61.7	57.85	71.06	85	77.72
CLIPping2025.12		99.88	96.74	94.77	94.87	99.47	94.15	76.48	60.27	61.26	60.78	80.77	80.69	53.95	63.94	84.76	80.19
Lgrad2025.12		99.8	86.94	85.63	91.08	99.27	72.49	56.42	44.47	63.03	63.67	67.1	66.1	56.2	63.6	67.02	72.19
NPR2025.12		99.8	96.1	84.4	97.7	99.3	82.5	80.2	69.2	76.6	77.9	69.7	77.3	77.8	76.1	78.1	82.85
FreDect2025.12		99.36	78.76	81.97	78.01	94.62	80.57	63.29	50	40.02	40.38	64.66	55.43	46.89	41.54	78.95	66.3
CausalCLIP2025.12		99.06	99.95	91.73	92.73	96.14	95.92	72.34	79.45	82.58	82.54	78.94	82.73	65.99	85.91	87.48	86.23