ResearchBenchmarksMassive Multitask Language Understanding on MMLU Full TextFollow64.9Normalized Log AccuracyT-Free62.19662.89863.664.302Mar 16, 2026Evaluation ResultsMethodMethodLinksNormalized Log AccuracyMemory (Bytes/Seq Pos)T-Freeshots=5-shotshots=5-shot2026.0364.95.31HATifiedshots=5-shotshots=5-shot2026.0363.85.31Llamashots=5-shotshots=5-shot2026.0362.34.56