@article{dynamicmemorycompressionretrofittingllms, title = {Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference}, author = {Piotr Nawrot and Adrian Łańcucki and Marcin Chochowski and David Tarjan and Edoardo M. Ponti}, year = {2024}, eprint = {2403.09636}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2403.09636v2}, }