ResearchDatasetsGPT-2 Concept SteeringGPT-2 Concept SteeringFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsConcept SteeringGPT-2 Concept Steering 3 concepts x 4 layers x 25 contexts (test)1.32Median Off-target KL Ratio12
Concept SteeringGPT-2 Concept Steering 3 concepts x 4 layers x 25 contexts (test)1.32Median Off-target KL Ratio12