Mixtures of In-Context Learners: A Robust AI Solution for Managing Memory Constraints and Improving Classification Accuracy in Transformer-Based NLP Models