TY - RPRT TI - VL-Cache: Sparsity and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration AU - Dezhan Tu AU - Danylo Vashchilenko AU - Yuzhe Lu AU - Panpan Xu PY - 2024 UR - https://arxiv.org/abs/2410.23317 ID - 2410.23317 ER -