Model Accuracy Drops After Batch Size Increase in TensorFlow – Why?

hanged batch size from 32 to 128, and validation accuracy dropped. No other hyperparameters were modified. Anyone experienced similar issues or have tuning advice?