KV Cache Compression Through the Lens of Transform Coding
Attention-Aware Transform Coding is introduced, which allocates bits over a calibration set to minimize attention-aware distortion and achieves near-lossless accuracy at approximately $5.8\times compression.