💻 computer science
Squeezing the Cache, Preserving the Truth: Monotonic Equipotential Allocation with Geodesia-KV
Geodesia-KV is a training-free, hardware-ready KV cache compression framework that employs monotonic block-wise precision allocation and query-sparse reading to significantly outperform existing methods in perplexity and bandwidth efficiency, enabling 1M-token context generation on consumer-grade 16 GiB GPUs.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.