HeadWiseKV: Budgeted Per-Head Cache Residency for Hybrid Long-Context Language Models

Overview

arXiv:2609.02029v1 Announce Type: new Abstract: Long-context inference retains a growing key--value (KV) cache during decoding, which consumes substantial GPU memory and can reduce generation throughput. This bottleneck remains in hybrid language models because their residual global-attention layers can dominate context-dependent cache demand. We study how to allocate this state under an aggregate KV-residency budget. We introduce HeadWiseKV, a training-free framework that compresses the residua

SpinGraph analysis pending — check back after processing.