Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Multi-task Language Understanding on MMLU (Adversarial Robustness)
Loading...
100
Accuracy under Attack
Reporting-and-penalty mechanism
40.72
56.11
71.5
86.89
Apr 26, 2026
Accuracy under Attack
Accuracy after Defense Recovery
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy under Attack
Accuracy after Defense Recovery
Reporting-and-penalty mechanism
Attack Type=All Attack...
2026.04
100
100
SentinelNet
Attack Type=Collaborat...
2026.04
76
93
GroupGuard
Attack Type=False Cons...
2026.04
52
92
GroupGuard
Attack Type=False Cons...
2026.04
44
97
GroupGuard
Attack Type=False Cons...
2026.04
43
84
Feedback
Search any
task
Search any
task