Fine-Tune an SLM or Prompt an LLM? The Case of Generating Low-Code Workflows

TL;DR

This paper compares fine-tuning small language models and prompting large language models for generating low-code workflows, finding fine-tuning still offers quality advantages despite the popularity of prompting LLMs.

Contribution

It provides empirical evidence that fine-tuning SLMs can outperform prompting LLMs in domain-specific structured output tasks, specifically low-code workflow generation.

Findings

Fine-tuning SLMs improves quality by 10% over prompting LLMs.

Prompting yields reasonable but less accurate results.

Error analysis reveals specific model limitations.

Abstract

Large Language Models (LLMs) such as GPT-4o can handle a wide range of complex tasks with the right prompt. As per token costs are reduced, the advantages of fine-tuning Small Language Models (SLMs) for real-world applications -- faster inference, lower costs -- may no longer be clear. In this work, we present evidence that, for domain-specific tasks that require structured outputs, SLMs still have a quality advantage. We compare fine-tuning an SLM against prompting LLMs on the task of generating low-code workflows in JSON form. We observe that while a good prompt can yield reasonable results, fine-tuning improves quality by 10% on average. We also perform systematic error analysis to reveal model limitations.