1 paper · 1 filter
Xiaoxin Shi, Jiaxin Wan, Linkang Dong +3
LLM-based function calling enables intelligent agents to interact with external tools and environments, yet autoregressive decoding imposes a fundamental latency bottleneck that li…