StateFlow: Enhancing LLM Task-Solving through State-Driven Workflows
Research Poster Engineering 2025 Graduate ExhibitionPresentation by Yiran Wu
Exhibition Number 46
Abstract
It is a notable trend to use Large Language Models (LLMs) to tackle complex tasks, e.g., tasks that require a sequence of actions and dynamic interaction with tools and external environments. In this paper, we propose StateFlow, a novel LLM-based task-solving paradigm that conceptualizes complex task-solving processes as state machines. In StateFlow, we distinguish between "process grounding” (via state and state transitions) and "sub-task solving” (through actions within a state), enhancing control and interpretability of the task-solving procedure. A state represents the status of a running process. The transitions between states are controlled by heuristic rules or decisions made by the LLM, allowing for a dynamic and adaptive progression. Upon entering a state, a series of actions is executed, involving not only calling LLMs guided by different prompts, but also the utilization of external tools as needed. Our results show that StateFlow significantly enhances LLMs' efficiency. For instance, StateFlow achieves 13% and 28% higher success rates compared to ReAct in InterCode SQL and ALFWorld benchmark, with 5x and 3x less cost respectively. We also show that StateFlow can be combined with iterative refining methods like Reflexion to further improve performance.
Importance
Large language models (LLMs) are powerful tools for solving complex tasks, but they often lack precise control over multi-step workflows. This can lead to inefficiency and errors in decision-making. Our study introduces StateFlow, a framework that uses finite state machines to guide LLM workflows with clear steps, improving control and reducing resource usage. By breaking tasks into defined states and automating transitions between them, StateFlow ensures tasks are completed more effectively and efficiently. Tested on coding and virtual environment tasks, StateFlow outperformed existing methods, achieving higher success rates with significantly lower costs. This approach offers a practical solution for enhancing LLM reliability and has the potential to improve applications across industries requiring complex decision-making systems.