A curated list of papers on workflow optimization for LLM agents.
This repository accompanies the survey paper: From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agents (https://arxiv.org/abs/2603.22386).
This list focuses on papers whose main contribution is workflow optimization. To keep the repository concise and method-centric, it does not aim to cover general survey papers, benchmarks, datasets, or framework/tooling papers.
The list is organized by when workflow structure is determined: static optimization and dynamic optimization.
| Title | Year | Paper |
|---|---|---|
| AFlow: Automating Agentic Workflow Generation | 2025 | Link |
| Automated Design of Agentic Systems | 2025 | Link |
| A²Flow: Automating Agentic Workflow Generation via Self-Adaptive Abstraction Operators | 2026 | Link |
| SEW: Self-Evolving Agentic Workflows for Automated Code Generation | 2025 | Link |
| Evolutionary Generation of Multi-Agent Systems | 2026 | Link |
| Title | Year | Paper |
|---|---|---|
| DSPy: Compiling Declarative Language Model Calls into State-of-the-Art Pipelines | 2024 | Link |
| Large Language Models as Optimizers | 2024 | Link |
| Connecting Large Language Models with Evolutionary Algorithms Yields Powerful Prompt Optimizers | 2024 | Link |
| CAPO: Cost-Aware Prompt Optimization | 2025 | Link |
| GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning | 2026 | Link |
| Optima: Optimizing Effectiveness and Efficiency for LLM-Based Multi-Agent System | 2025 | Link |
| Title | Year | Paper |
|---|---|---|
| Multi-Agent Design: Optimizing Agents with Better Prompts and Topologies | 2026 | Link |
| Learning Multi-Agent Communication from Graph Modeling Perspective | 2024 | Link |
| Maestro: Joint Graph & Config Optimization for Reliable AI Agents | 2025 | Link |
| Title | Year | Paper |
|---|---|---|
| MermaidFlow: Redefining Agentic Workflow Generation via Safety-Constrained Evolutionary Programming | 2025 | Link |
| VFlow: Discovering Optimal Agentic Workflows for Verilog Generation | 2026 | Link |
| Title | Year | Paper |
|---|---|---|
| Cut the Crap: An Economical Communication Pipeline for LLM-Based Multi-Agent Systems | 2025 | Link |
| Adaptive Graph Pruning for Multi-Agent Communication | 2025 | Link |
| DAGP: Difficulty-Aware Graph Pruning for LLM-Based Multi-Agent Systems | 2025 | Link |
| AgentDropout: Dynamic Agent Elimination for Token-Efficient and High-Performance LLM-Based Multi-Agent Collaboration | 2025 | Link |
| A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration | 2024 | Link |
| MasRouter: Learning to Route LLMs for Multi-Agent Systems | 2025 | Link |
| SkillOrchestra: Learning to Route Agents via Skill Transfer | 2026 | Link |
| Title | Year | Paper |
|---|---|---|
| Difficulty-Aware Agentic Orchestration for Query-Specific Multi-Agent Workflows | 2026 | Link |
| Assemble Your Crew: Automatic Multi-Agent Communication Topology Design via Autoregressive Graph Generation | 2026 | Link |
| G-Designer: Architecting Multi-Agent Communication Topologies via Graph Neural Networks | 2025 | Link |
| Dynamic Generation of Multi-LLM Agents Communication Topologies with Graph Diffusion Models | 2025 | Link |
| Multi-Agent Architecture Search via Agentic Supernet | 2025 | Link |
| ScoreFlow: Mastering LLM Agent Workflows via Score-Based Preference Optimization | 2025 | Link |
| FlowReasoner: Reinforcing Query-Level Meta-Agents | 2025 | Link |
| Workflow-R1: Group Sub-sequence Policy Optimization for Multi-Turn Workflow Construction | 2026 | Link |
| AutoFlow: Automated Workflow Generation for Large Language Model Agents | 2024 | Link |
| WorkflowLLM: Enhancing Workflow Orchestration Capability of Large Language Models | 2025 | Link |
| RobustFlow: Towards Robust Agentic Workflow Generation | 2025 | Link |
| ComfyUI-R1: Exploring Reasoning Models for Workflow Generation | 2025 | Link |
| AutoAgents: A Framework for Automatic Agent Generation | 2024 | Link |
| Title | Year | Paper |
|---|---|---|
| DyFlow: Dynamic Workflow Framework for Agentic Reasoning | 2025 | Link |
| AgentConductor: Topology Evolution for Multi-Agent Competition-Level Code Generation | 2026 | Link |
| Aime: Towards Fully-Autonomous Multi-Agent Framework | 2025 | Link |
| AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration | 2026 | Link |
| MetaGen: Self-Evolving Roles and Topologies for Multi-Agent LLM Reasoning | 2026 | Link |
| ProAgent: From Robotic Process Automation to Agentic Process Automation | 2023 | Link |
| Flow: Modularized Agentic Workflow Automation | 2025 | Link |
| EvoFlow: Evolving Diverse Agentic Workflows on the Fly | 2025 | Link |
| DebFlow: Automating Agent Creation via Agent Debate | 2025 | Link |
| QualityFlow: An Agentic Workflow for Program Synthesis Controlled by LLM Quality Checks | 2025 | Link |
Contributions are welcome! To suggest a new paper, please open an issue with the following information:
- Paper title, year, and a link to the official paper page.
- The subsection you think it belongs to.
- A brief reason why its main contribution is workflow optimization.
We will review and update the list accordingly. Please note:
- Only papers whose main contribution is workflow optimization will be included.
- Each paper is placed in exactly one subsection that best matches its main contribution.
- We prefer official paper pages (e.g., conference proceedings, arXiv).
If you find this repository or the accompanying survey helpful, please cite:
@article{yue2026survey,
title={From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agents},
author={Yue, Ling and Bhandari, Kushal Raj and Ko, Ching-Yun and Patel, Dhaval and Lin, Shuxin and Zhou, Nianjun and Gao, Jianxi and Chen, Pin-Yu and Pan, Shaowu},
journal={arXiv preprint arXiv:2603.22386},
year={2026}
}