List your company

Poisoned Examples Removal on HellaSwag

93.7AccuracyOracle Exp.
88.18889.61991.0592.481May 18, 2025

Evaluation results

MethodLinks
Oracle Exp.Model=GPT-4o2025.0593.7
Proxy Exp.Model=GPT-4o2025.0593.5
Random deletionModel=GPT-4o2025.0588.4