Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Pooled Cross-Model Evaluation Suite

Benchmarks

Task NameDataset NameSOTA ResultTrend
Consequential-Action Attack Success RatePooled Cross-Model Evaluation Suite (all split)
Attack Success Rate (ASR)31.2
5
Consequential-Action Attack Success RatePooled Cross-Model Evaluation Suite whitebox
ASR22.2
5
Consequential-Action Attack Success RatePooled Cross-Model Evaluation Suite (adaptive split)
Attack Success Rate (ASR)27.6
5
Consequential-Action Attack Success RatePooled Cross-Model Evaluation Suite static split
ASR (%)43.8
5
Utility evaluationPooled Cross-Model Evaluation Suite
Utility95.9
4
Showing 5 of 5 rows