Toward Reliable Design of LLM-Enabled Agentic Workflows: Optimizing Latency-Reliability-Cost Tradeoffs
Research area: Machine Learning Authors: Ya-Ting Yang, Quanyan Zhu arXiv: 2505.21640
Abstract
Modern AI systems increasingly rely on workflows composed of multiple interacting agents, some powered by large language models (LLMs) and others by conventional computational modules. This paper analyzes the fundamental tradeoffs between latency, reliability, and cost in LLM-enabled agentic workflows. The authors introduce performance models for both LLM and non-LLM agents that capture the relationship between computational effort and output quality, incorporating the impact of reasoning and output tokens for LLM agents using a parametric exponential reliability function. Then, they study the design of sequential workflows under latency and cost constraints. Main results include a water-filling token allocation policy and characterizations of optimal workflow reliability in terms of shadow prices.
Key points
- Analyzes latency, reliability, and cost tradeoffs in workflows mixing LLM-powered and conventional agents.
- Proposes performance models relating computational effort to output quality for both LLM and non-LLM agents.
- Uses a parametric exponential reliability function to capture the effect of reasoning and output tokens on LLM agents.
- Derives a water-filling token allocation policy for sequential workflow design under latency and cost constraints.
- Characterizes optimal workflow reliability via shadow prices.