Skip to content

WorkflowLLM

A data-centric framework for workflow orchestration

Visit official site

What WorkflowLLM is

The framework is described as a data-centric effort to strengthen workflow orchestration in large language models. It centres on WorkflowBench and finetunes Llama-3.1-8B into WorkflowLlama for orchestration tasks.

What you can do with it

  • Install Python 3.8 and the listed dependencies
  • Run the shortcut conversion preprocessing script
  • Train with train.sh and run inference with infer.sh

Application scenarios

  • Developers automate software development flows
  • Data scientists handle large-scale datasets
  • Automation engineers design industrial flows

Pros & cons

✓ What we like

  • Query expansion with ChatGPT enriches the dataset
  • plist shortcuts are converted into AST representations

! What to watch out for

  • Setup calls for Python 3.8 and requirements.txt
  • Training takes base model and data paths as arguments

FAQ

How large is WorkflowBench?

The material states 106763 samples, 1503 APIs, 83 apps and 28 categories.

Which model is finetuned?

Llama-3.1-8B is finetuned to create WorkflowLlama.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review