2 ms·MSFT's small language model outperforms larger models on standardized math tests3 points by Brajeshwar 3y ago