作者位次 #8/9NeurIPS 2026Poster· Deep Learning / Attention Mechanisms
Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse
Taekyung Heo · Rasoul Shafipour · Ritchie Zhao · Maximilian Golub · Mohammad Mahdi Kamani · Ritika Borkar · Makesh T Chandran · Pantea Zardoshti · Bita Darvish Rouhani