Story
arxiv_cs_lg · Jul 6, 2026 · paper
Source brief
정책 경사를 활용한 적응형 추론 배치
arxiv.orgJul 6, 2026
original source linked
In brief
Inference serving systems must balance throughput and latency under bursty, heterogeneous workloads, yet the industry standard remains static batching policies that require manual tuning and cannot adapt to shifting t...
Feed lens
agenteval