Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2508.08053 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909732452696064 |
|---|---|
| author | Zhu, Runchuan Jiang, Bowen Mei, Lingrui Yang, Fangkai Wang, Lu Gao, Haoxiang Bai, Fengshuo Zhao, Pu Lin, Qingwei Rajmohan, Saravan Zhang, Dongmei |
| author_facet | Zhu, Runchuan Jiang, Bowen Mei, Lingrui Yang, Fangkai Wang, Lu Gao, Haoxiang Bai, Fengshuo Zhao, Pu Lin, Qingwei Rajmohan, Saravan Zhang, Dongmei |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_08053 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AdaptFlow: Adaptive Workflow Optimization via Meta-Learning Zhu, Runchuan Jiang, Bowen Mei, Lingrui Yang, Fangkai Wang, Lu Gao, Haoxiang Bai, Fengshuo Zhao, Pu Lin, Qingwei Rajmohan, Saravan Zhang, Dongmei Artificial Intelligence 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. |
| title | AdaptFlow: Adaptive Workflow Optimization via Meta-Learning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2508.08053 |