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

Sample-level Membership Inference on LIBERO Long

1AUCAgg.-NLL

Evaluation Results

Method	Links
Agg.-NLL2026.05		1	100	100	100
Agg.-Conf2026.05		1	100	100	100
Temp.-Smooth2026.05		1	100	100	100
Temp.-Curve2026.05		1	100	100	100
Action-L12026.05		1	100	100	100
Action-MSE2026.05		1	100	100	100
Agg.-NLL2026.05		0.9996	98	98	100
NLL2026.05		0.9995	88.78	99.15	99.97
Action-L12026.05		0.9612	96	96	96
Action-L12026.05		0.952	30.16	63.78	85.12
Action-L12026.05		0.9435	63.88	74.9	84.49
Action-MSE2026.05		0.941	64.94	75.97	85.11
Action-MSE2026.05		0.9387	93	93	93
Action-MSE2026.05		0.9254	25.08	61.62	82.62
Conf2026.05		0.8811	10.68	30.13	52.94
NLL2026.05		0.7344	0.92	7	21.47
Temp.-Curve2026.05		0.6716	1	2	15
Agg.-Conf2026.05		0.6462	10	12	19
Temp.-Smooth2026.05		0.61	0	0	8
Conf2026.05		0.5518	0.86	4.37	11.85
Agg.-Conf-Fix2026.05		0.5489	3	3	10
Conf-FixPrompt2026.05		0.5007	0.09	0.87	5
Conf-FixPrompt2026.05		0.4958	0.21	1.5	6.39
Agg.-Conf-Fix2026.05		0.4839	0	0	10