ResearchBenchmarksInterpretability Method Evaluation on Interpretability Method ComparisonInterpretability Method Evaluation on Interpretability Method ComparisonFollow2CostPert. probing1.91.9522.05Apr 30, 2026Evaluation resultsMethodMethodLinksCostCausal ScoreTask SpecificityTrain-Free ScoreDirect Injection ScorePert. probing2026.042————