arXivTing-Yun Chang, Harvey Yiyun Fu, Deqing Fu, Chenghao Yang, Jesse Thomason, Robin JiaTue, Jun 2, 2026, 10:16 AM PDT
score 16.5
Smarter cache deletion cuts reasoning model memory by 4x
Original: Value-Aware Stochastic KV Cache Eviction for Reasoning Models
Source: arxiv.org ↗
Writing ELI5 summary…