ResearchBenchmarksTruthfulness on TruthfulQA DE (test)Follow34.2Normalized Probability MassLlama33.99234.04634.134.154Mar 16, 2026Evaluation ResultsMethodMethodLinksNormalized Probability MassBytes per Sequence PositionLlamashots=6-shot, evaluati...shots=6-shot, evaluation_mode=Pre-training2026.0334.23.39HATifiedshots=6-shot, evaluati...shots=6-shot, evaluation_mode=Pre-training2026.0334.15.91T-Freeshots=6-shot, evaluati...shots=6-shot, evaluation_mode=Pre-training2026.03345.91