KV Cache and Inference Efficiency

Papers on reducing memory and computational overhead during inference through KV cache optimization, prefix caching, and efficient decoding strategies for language and protein models.