Surface-form evasion probe suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Surface-form evasion probe suite Restoration-trigger injection
58.8Exposure Rate
2
Surface-form evasion probe suite Mixed-language mentions
38.8Exposure Rate
2
Surface-form evasion probe suite Paraphrase-sensitive spans
47.6Exposure
2
Surface-form evasion probe suite Homoglyph substitution
43.9Exposure Rate
2