1 paper · 2 filters
Yifei Wang, Hancheng Ye, Yechen Xu +8
Large language models (LLMs) are increasingly deployed as the execution core of autonomous agents rather than as standalone text generators. Agentic workloads induce a temporal shi…