Jul 2026
ResKV: Reconstructing Omitted Attention Contributions for Fixed-Budget KV Cache Compression
ResKV is proposed, which divides a fixed KV budget into an exact main cache and a compact residual cache that reconstructs the contribution of omitted tokens and lets main-cache tokens and residual entries participate in the same softmax normalization.
Yuhan Zhan, Lisi Chen, Shuo Shang
· arXiv.org · 2 citations