Deepfake Detection on HydraFake benchmark leaderboard · SOTA2 Research

Deepfake Detection on HydraFake

98.2ID AccuracyVIGIL

Evaluation Results

Method	Links
VIGIL2026.03		98.2	96.3	99.5	96.2	97.2	99.1	99.6	95.5	88	97.1	94.9	95.9	93.4	67.5	86.9	94.1	91.5	85.7	73.5	92.1
Veritas2026.03		97.3	94.8	99.8	97	99.9	99.9	99.9	90.3	75.7	97	91.8	95.1	91.7	58.6	84.1	92.3	90.2	89.2	78.5	90.7
DDA2026.03		95.3	81.7	97.2	95.3	96.6	97	96.5	94.1	91.8	96.8	94.7	82.9	95.9	63.9	85.8	64.9	67.1	52.4	79.3	85.7
D^32026.03		93.4	96.7	98.3	93.3	97.1	99.2	98.2	66.3	73.2	93.6	80.2	81	90.1	78.9	71.4	92.2	86.2	85.7	70.2	86.6
Co-SPY2026.03		88.9	89.8	94.2	82.8	92.1	94.5	93.3	81	87.7	90.1	89.1	76.1	94.2	73.3	84.2	72.6	79.2	80.8	66.2	84.7
C2P-CLIP2026.03		87.3	66.2	75.5	85	79.4	82.2	80.5	65	67.1	81.9	74.6	76.7	88.2	59.7	70.3	72.2	69.6	51.8	69.6	73.8
NPR2026.03		86.4	65.2	95	62.3	97	85.3	83.2	48.3	52.3	52.1	89.5	87.5	66.7	55.5	69.7	81.6	59	53.7	54.2	70.8
UnivFD2026.03		83.2	89	93	81.6	73.7	95.1	92.1	60.8	81.6	74	72.9	68.1	79.8	66.6	67.7	80.6	74.4	73.6	68.3	77.7
FakeVLM2026.03		78.5	78.2	78.5	77	74.5	76.5	76.8	70.8	76.2	76.2	76.9	76.5	77.7	75.7	83.6	81.5	80.8	78.7	74.5	77.3
AIDE2026.03		77.3	84	81.3	88	81.3	85.1	64.7	54.4	73.9	86.1	77.8	59.9	78.2	54.3	59.7	64.9	58	50.8	51.9	70.1
SAFE2026.03		73.4	63.3	73.8	69.8	72.3	99.4	63.6	49.9	97.4	50.5	50.2	74.1	53.4	59.9	51.4	83.5	54.9	50.1	49.7	65.3
GPT-5.22026.03		71.9	90.6	83.6	67.4	57.8	86.2	74.8	78.9	67.6	64.3	75.6	66.3	71.8	61.1	73.7	51.5	80.6	25.9	53.9	68.6
Gemini-3-Pro2026.03		70	87.5	87.8	75.9	49.7	84	87.2	76.3	63	69.9	79.6	80.3	75.3	62.8	78.3	85.5	79.3	41.8	53.1	73
Qwen3-VL-8B2026.03		64.8	57.6	56.3	51.2	49.3	80	62.8	53.6	53.9	55	53.9	57.1	55.9	51.8	56.9	69.6	71.2	8.8	54.3	56
Kimi-K2.52026.03		64.3	91.2	81.9	64.5	56	88.5	77	60.4	64.6	68.6	60.1	61.8	71.2	62.9	75.9	95.2	78	60.9	53.3	70.3
GLM-4.6V2026.03		60	55	50.1	51.5	50.5	57.4	51.4	54.1	49.5	53.7	52.9	55.7	50.1	53.7	71.7	59.6	75.5	18.7	51.1	53.8