RLVR Data Contamination Detection on RL-MIA various RL-tuned models 1.0 benchmark leaderboard · SOTA2 Research

RLVR Data Contamination Detection on RL-MIA various RL-tuned models 1.0

0.72DAPOMin-kNN Distance

Evaluation Results

Method	Links
Min-kNN Distance2026.02		0.72	0.66	0.75	0.76	0.8	0.68	0.68	0.57	0.7
PPL2026.02		0.63	0.61	0.72	0.72	0.59	0.56	0.44	0.54	0.6
Self-Critique2026.02		0.58	0.51	0.48	0.49	0.54	0.53	0.43	0.49	0.51
CDD2026.02		0.52	0.51	0.36	0.45	0.49	0.46	0.5	0.47	0.47
Recall2026.02		0.5	0.52	0.51	0.62	0.54	0.52	0.56	0.47	0.53
Min-K%2026.02		0.37	0.43	0.29	0.28	0.53	0.44	0.53	0.47	0.42
Min-K%++2026.02		0.37	0.48	0.56	0.39	0.56	0.53	0.66	0.52	0.51