Dataset

Wiki

XSum

AG News

Metrics

Accuracy

Precision

Recall

Accuracy

Precision

Recall

Accuracy

Precision

Recall

Neighborhood Attack

Mistral 7B

75.3

67.6

97.0

86.1

78.2

100

75.8

67.4

100

LLaMA 7B

74.9

66.6

100

83.4

75.1

100

77.7

69.2

100

LLaMA 13B

75.3

66.9

100

83.7

75.4

100

76.4

67.9

100

Mixtral 8x7B

77.3

68.8

100

85.7

77.8

99.8

76.6

68.2

99.8