Mixture of Scales: Memory-Efficient Token-Adaptive Binarization for Large Language Models

Re-assign community