ResearchDatasetsGSM8K ⊕ BeaverTailsGSM8K ⊕ BeaverTailsFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsEmbedded Direct AttackGSM8K ⊕ BeaverTails74.7Utility12Backdoor AttackGSM8K ⊕ BeaverTails10ASR (No Trigger)8