llm-substitution

This paper replaces LLM-based decision making in edge service orchestration with Jev decision models, reporting 22.7–64.5% lower median decision latency and far higher request completion rates under load, while noting wide contracts mark the limit of substitution.