Sample-level Membership Inference on LIBERO Goal benchmark leaderboard · SOTA2 Research

Sample-level Membership Inference on LIBERO Goal

100AUCAgg.-NLL

Evaluation Results

Method	Links
Agg.-NLL2026.05		100	100	100	100
Agg.-NLL2026.05		100	100	100	100
Agg.-Conf2026.05		100	100	100	100
Action-L12026.05		100	100	100	100
NLL2026.05		99.99	99.62	99.98	100
Action-MSE2026.05		99.93	98	99	99
Temp.-Curve2026.05		99.79	93	97	99
Agg.-Conf2026.05		99.71	95	96	98
Temp.-Smooth2026.05		99.68	92	94	98
Action-L12026.05		98.54	21.24	87.22	96.01
Action-MSE2026.05		98.03	24.57	84.6	95.33
Action-L12026.05		96.15	96	96	96
Action-MSE2026.05		96.07	93	95	96
Action-L12026.05		94.81	10.62	47.66	76.41
Action-L12026.05		92.07	52.5	69.9	79.36
Action-MSE2026.05		91.84	44.86	70.35	79.8
Action-MSE2026.05		90.64	12.3	49.72	72.86
Flow-Loss2026.05		89.14	8.41	30.23	52.87
Conf2026.05		84.1	4.94	17.12	42.29
NLL2026.05		79.27	1.14	8.35	27.55
Conf2026.05		78.96	3.62	14.8	35.82
Temp.-Curve2026.05		74.75	15	15	21
Agg.-Conf-Fix2026.05		72.52	5	16	32
Temp.-Smooth2026.05		65.83	9	9	16
Conf-FixPrompt2026.05		56.18	0.5	3.01	9.64
Flow-Loss-FixPrompt2026.05		52.6	0.01	1.01	5.39
Conf-FixPrompt2026.05		50.06	0.1	0.82	4.67
Agg.-Conf-Fix2026.05		47.74	0	4	7