ResearchBenchmarksTruthfulness on TruthfulQA (norm. prob. mass and compression)Follow36.4Normalized Probability MassT-Free34.94435.32235.736.078Mar 16, 2026Evaluation ResultsMethodMethodLinksNormalized Probability MassCompression ScoreT-Freeshots=6-shotshots=6-shot2026.0336.44.91HATifiedshots=6-shotshots=6-shot2026.0335.74.91Llamashots=6-shotshots=6-shot2026.03354.18Tülushots=6-shotshots=6-shot2026.03354.18