ResearchBenchmarksQuestion Answering on MMLU Health Large Shift (test)Follow50.6AccuracyBayesian-LoRA47.16848.05948.9549.841Jan 28, 2026Evaluation ResultsMethodMethodLinksAccuracyECENLLBayesian-LoRASample size (S)=4, Tra...Sample size (S)=4, Training protocol=End-to-End2026.0150.612.91.2LA2026.0149.117.61.35LLLA2026.0148.533.51.31MAP2026.0148.334.22.09Dropout2026.0148.233.52.05Ckpt Ens2026.0148.2291.74Temp2026.0147.717.41.36BBB2026.0147.627.41.29BLoBEnsemble size (N)=10Ensemble size (N)=102026.0147.316.41.94