ResearchBenchmarksCommonsense Reasoning on PIQA (Accuracy %)Follow61.1Accuracy (%)LoopMDM58.39659.09859.860.502May 25, 2026Evaluation ResultsMethodMethodLinksAccuracy (%)LoopMDMS=12, Zero-shot=trueS=12, Zero-shot=true2026.0561.1MDMZero-shot=trueZero-shot=true2026.0560.6LoopMDMS=6, Zero-shot=trueS=6, Zero-shot=true2026.0558.9LoopMDMS=1, Zero-shot=trueS=1, Zero-shot=true2026.0558.5