HARD-KV: Head-Adaptive Regularization for Decoding-time KV Compression

arXiv:2606.28831v1 Announce Type: new Abstract: Long-context LLM inference faces a fundamental conflict: head-adaptive compression algorithms (e.g., Top-$p$ nucleus sampling) offer superior accuracy by dynamically fluctuating memory budgets, yet modern inference engines (e.g., vLLM) demand rigid, static memory patterns to leverage CUDA Graphs and PagedAttention. We resolve this ``Static-Dynamic'' mismatch with HARD-KV, a unified framework that that bridges dynamic selection with rigid system con...