ResearchBenchmarksTargeted Failure Generation on DrivingFollow0.102Diversity TGIFTbench0.093680.095840.0980.10016Jan 21, 2026Evaluation ResultsMethodMethodLinksDiversity TMisclass RateEscape RatioConfidence ReductionMS-SSIMLPIPSTrace DiffGIFTbench2026.010.102—————191.5HyNeA2026.010.094—————60.71