AdaptFlow: Adaptive Workflow Optimization via Meta-Learning

TL;DR

AdaptFlow introduces a meta-learning framework that enables large language models to rapidly adapt workflows to new tasks using language-guided modifications, improving flexibility and performance across diverse benchmarks.

Contribution

It presents a novel meta-learning approach inspired by MAML for workflow adaptation in LLMs, enabling quick task-specific customization through language feedback.

Findings

Outperforms manual and searched workflows on multiple benchmarks.

Achieves state-of-the-art results with strong task generalization.

Effective across question answering, code generation, and reasoning tasks.

Abstract

Recent advances in large language models (LLMs) have sparked growing interest in agentic workflows, which are structured sequences of LLM invocations intended to solve complex tasks. However, existing approaches often rely on static templates or manually designed workflows, which limit adaptability to diverse tasks and hinder scalability. We propose AdaptFlow, a natural language-based meta-learning framework inspired by model-agnostic meta-learning (MAML). AdaptFlow learns a generalizable workflow initialization that enables rapid subtask-level adaptation. It employs a bi-level optimization scheme: the inner loop refines the workflow for a specific subtask using LLM-generated feedback, while the outer loop updates the shared initialization to perform well across tasks. This setup allows AdaptFlow to generalize effectively to unseen tasks by adapting the initialized workflow through language-guided modifications. Evaluated across question answering, code generation, and mathematical reasoning benchmarks, AdaptFlow consistently outperforms both manually crafted and automatically searched baselines, achieving state-of-the-art results with strong generalization across tasks and models. The source code and data are available at https://github.com/microsoft/DKI_LLM/tree/AdaptFlow/AdaptFlow.