ResearchBenchmarksModel Distinguishability on Language Model DatasetModel Distinguishability on Language Model DatasetFollow0.083Directional Sharpness RatioLM-Mem0.05180.05990.0680.0761Jun 23, 2026Evaluation resultsMethodMethodLinksDirectional Sharpness RatioStatic Sharpness RatioObserved Accuracy RatioLM-MemExperiment=Language Mo...Experiment=Language Model Memorization2026.060.0830.6641.003LM-BDExperiment=Language Mo...Experiment=Language Model Backdoor2026.060.0530.631.005