ResearchBenchmarksGraduate-Level Google-Proof Q&A on GPQAGraduate-Level Google-Proof Q&A on GPQAFollow0.308Log AccuracyHATified0.27160.281050.29050.29995Mar 16, 2026Evaluation resultsMethodMethodLinksLog AccuracyBytes per Sequence PositionHATifiedshots=0-shotshots=0-shot2026.030.3084.93Llamashots=0-shotshots=0-shot2026.030.3013.52T-Freeshots=0-shotshots=0-shot2026.030.2734.93