# Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai

Source: [Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai](<https://www.sota2.com/research/paper/68fa75d31e209131cea6a125>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T04:59:38.808Z

Authors: Parinthapat Pengpun, Can Udomcharoenchaikit, Weerayut Buaphet, Peerat Limkonchotiwat

Published: 2024-11-23T07:50:59.000Z

## Abstract

We present a synthetic data approach for instruction-tuning large language models (LLMs) for low-resource languages in a data-efficient manner, specifically focusing on Thai. We identify three key properties that contribute to the effectiveness of instruction-tuning datasets: fluency, diversity, and cultural context. We propose a seed-data-free framework for generating synthetic instruction-tuning data that incorporates these essential properties. Our framework employs an LLM to generate diverse topics, retrieve relevant contexts from Wikipedia, and create instructions for various tasks, such as question answering, summarization, and conversation. The experimental results show that our best-performing synthetic dataset, which incorporates all three key properties, achieves competitive performance using only 5,000 instructions when compared to state-of-the-art Thai LLMs trained on hundreds of thousands of instructions. Our code and dataset are publicly available at https://github.com/parinzee/seed-free-synthetic-instruct.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Average evaluation across 7 tasks on Thai Culture (test)](<https://www.sota2.com/research/sota/average-evaluation-across-7-tasks-on-thai-culture-test.md>): Average evaluation across 7 tasks; Thai Culture (test); BERTScore: 69.5.

- [Average Evaluation across 7 Tasks on General (test)](<https://www.sota2.com/research/sota/average-evaluation-across-7-tasks-on-general-test.md>): Average evaluation across 7 tasks; General (test); BERTScore: 73.2.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2411.15484>)

[PDF](<https://arxiv.org/pdf/2411.15484.pdf>)

[Implementation](<https://github.com/parinzee/seed-free-synthetic-instruct>)
