FlowBot: Inducing LLM Workflows with Bilevel Optimization and Textual Gradients

LLM workflows, which coordinate structured calls to individual LLMs/agents to achieve a particular goal, offer a promising path towards building powerful AI systems that can tackle diverse tasks. However, existing approaches for building such workflows generally rely on human-crafted pipelines and prompts, which presents a substantial bottleneck in real world deployment. How can we automatically induce LLM-based agents and workflows in a data-driven way? This paper describes a simple data-driven approach for automatically inducing agents and LLM workflows. We formulate workflow induction as a bilevel optimization problem: an outer loop which optimizes a high-level sketch of the workflow (in particular how the LLM calls should be structured), and an inner loop which optimizes each individual LLM call one-by one. Both loops are optimized with "textual gradients" where for the inner loop we optimize each component in a modular way through "backpropagating" textual gradients layer-by-layer. We find that LLM workflows discovered through our FlowBot (workflow induction through bilevel optimization and textual gradients) approach performs competitively against strong baselines that make use of human-crafted or generated workflows.

Today's most capable AI systems rarely rely on a single model call. Instead, they often use multi-step workflows: one step looks up information, another checks it, and another writes the final answer. But building these workflows is still slow and manual. Experts must decide which steps are needed, how they should be connected, and what instructions each step should follow. We built FlowBot, a method that automatically designs such workflows from task examples. Like a student learning from written comments, FlowBot uses plain-language feedback to improve both the workflow's structure and the instructions for each step. On tasks including multi-step question answering, math reasoning, and code generation, FlowBot's automatically designed workflows match or outperform carefully hand-designed systems, while costing less to run. This points toward AI systems that can help build and refine their own problem-solving workflows.