AI-generated image detection on ForenSynth (test) benchmark leaderboard · SOTA2 Research

AI-generated image detection on ForenSynth (test)

98.6StyleGAN AccuracyDRIFT

Evaluation Results

Method	Links
DRIFT2026.06		98.6	100	94.3	99.3	99.1	100	92.8	100	91.2	97.8	97.8	99.8
DRIFT2026.06		98.6	100	94.3	99.3	99.1	100	92.8	100	91.2	97.8	97.8	99.8
SAFE2026.06		98	99.9	89.7	95.9	98.9	99.8	91.5	97.2	93.1	97.5	96.2	98.8
FerretNet2026.06		98	100	92.6	98.5	98.8	99.9	91.4	99.8	89.2	96.7	95.9	99.3
SAFE2026.06		98	99.9	89.7	95.9	98.9	99.8	91.5	97.2	93.1	97.5	96.2	98.8
FerretNet2026.06		98	100	92.6	98.5	98.8	99.9	91.4	99.8	89.2	96.7	95.9	99.3
FatFormer2026.06		97.2	99.8	99.5	100	99.3	100	99.4	100	93.2	98	98.4	99.7
FatFormer2026.06		97.2	99.8	99.5	100	99.3	100	99.4	100	93.2	98	98.4	99.7
NPR2026.06		96.3	99.8	87.5	94.5	95	99.5	86.6	88.8	77.4	86.2	92.5	96.1
NPR2026.06		96.3	99.8	87.5	94.5	95	99.5	86.6	88.8	77.4	86.2	92.5	96.1
LGrad2026.06		94.8	99.9	82.9	90.7	85.3	94	72.4	79.3	58	67.9	86.1	91.5
F3Net2026.06		92.6	99.7	65.3	69.9	76.4	84.3	58.1	56.7	63.5	78.8	80.4	86.2
FreqNet2026.06		90.2	99.7	90.5	96	95.8	99.6	93.4	98.6	88.9	94.4	91.5	98.5
FreqNet2026.06		90.2	99.7	90.5	96	95.8	99.6	93.4	98.6	88.9	94.4	91.5	98.5
Ojha2026.06		89	98.7	90.5	99.1	87.9	99.8	89.9	100	80.2	90.2	89.1	98.3
Ojha2026.06		89	98.7	90.5	99.1	87.9	99.8	89.9	100	80.2	90.2	89.1	98.3
FrePGAN2026.06		80.7	89.6	69.2	71.1	71.1	74.4	60.3	71.7	70.9	91.9	79.4	87.2
BiHPF2026.06		76.9	75.1	84.9	81.7	81.9	78.9	69.5	78.1	54.4	54.6	78.6	77.9
CO-SPY2026.06		63.9	70.2	71.6	83.9	58.5	55.8	69.6	83.4	65.7	79.7	65.7	76.1
CO-SPY2026.06		63.9	70.2	71.6	83.9	58.5	55.8	69.6	83.4	65.7	79.7	65.7	76.1
Wang2026.06		63.8	91.4	52.9	73.3	72.7	88.6	63.9	92.2	51.7	62.3	67.1	86.9