Face Identity Unlearning on VGGFace2 benchmark leaderboard · SOTA2 Research

Face Identity Unlearning on VGGFace2

63Original ForgettingRMU

Evaluation Results

Method	Links
RMU2026.05		63	68	70.33	67.43	91.67	82.86	73.88	62.33	63.67
Qwen3-VL-4B-Instruct2026.05		67.67	61.29	73.33	61.14	16.33	88.29	61.34	66	66.2
SLUG2026.05		80.33	48.71	84.33	48	29.67	66.71	59.63	78.67	27.47
NPO2026.05		91.67	52.29	90.67	52.57	71	77	72.53	99	63.67
UNDIAL2026.05		97	62.86	97.33	61	29.67	77.29	70.86	81.33	66.31
GA2026.05		97.33	22.57	98.33	18	96	4.14	56.06	0	62.68
GD2026.05		98	11.14	97	17	88.33	19.57	55.17	0	62.99
SimNPO2026.05		99	30.29	99	28.14	13.33	63.86	55.6	100	66.44
SAUCE2026.05		99	30.43	98.67	30.43	99.67	23.71	63.65	97.33	60.79
HFRU2026.05		99	67.29	98.67	78.14	24	83.86	75.16	66.67	66.28
HFRU2026.05		99.67	98.86	99.67	98.86	99.33	98.71	99.18	58.33	66.51
HFRU2026.05		99.67	99.86	99.33	99.57	99.67	99.86	99.66	1	66.51