Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tang, Shuo, Pang, Xianghe, Liu, Zexi, Tang, Bohan, Ye, Rui, Jin, Tian, Dong, Xiaowen, Wang, Yanfeng, Chen, Siheng
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910836416577536
author Tang, Shuo
Pang, Xianghe
Liu, Zexi
Tang, Bohan
Ye, Rui
Jin, Tian
Dong, Xiaowen
Wang, Yanfeng
Chen, Siheng
author_facet Tang, Shuo
Pang, Xianghe
Liu, Zexi
Tang, Bohan
Ye, Rui
Jin, Tian
Dong, Xiaowen
Wang, Yanfeng
Chen, Siheng
contents Post-training is essential for enabling large language models (LLMs) to follow human instructions. However, its effectiveness depends on high-quality instruction data, which is challenging to obtain in the real world due to privacy concerns, data scarcity, and high annotation costs. To fill this gap, inspired by the recent success of using LLMs to simulate human society, we propose MATRIX, a multi-agent simulator that automatically generates diverse text-based scenarios, capturing a wide range of real-world human needs in a realistic and scalable manner. Leveraging these outputs, we introduce a novel scenario-driven instruction generator MATRIX-Gen for controllable and highly realistic data synthesis. Extensive experiments demonstrate that our framework effectively generates both general and domain-specific data. On AlpacaEval 2 and Arena-Hard benchmarks, Llama-3-8B-Base, post-trained on datasets synthesized by MATRIX-Gen with just 20K instruction-response pairs, outperforms Meta's Llama-3-8B-Instruct model, which was trained on over 10M pairs.
format Preprint
id arxiv_https___arxiv_org_abs_2410_14251
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation
Tang, Shuo
Pang, Xianghe
Liu, Zexi
Tang, Bohan
Ye, Rui
Jin, Tian
Dong, Xiaowen
Wang, Yanfeng
Chen, Siheng
Artificial Intelligence
Computation and Language
Post-training is essential for enabling large language models (LLMs) to follow human instructions. However, its effectiveness depends on high-quality instruction data, which is challenging to obtain in the real world due to privacy concerns, data scarcity, and high annotation costs. To fill this gap, inspired by the recent success of using LLMs to simulate human society, we propose MATRIX, a multi-agent simulator that automatically generates diverse text-based scenarios, capturing a wide range of real-world human needs in a realistic and scalable manner. Leveraging these outputs, we introduce a novel scenario-driven instruction generator MATRIX-Gen for controllable and highly realistic data synthesis. Extensive experiments demonstrate that our framework effectively generates both general and domain-specific data. On AlpacaEval 2 and Arena-Hard benchmarks, Llama-3-8B-Base, post-trained on datasets synthesized by MATRIX-Gen with just 20K instruction-response pairs, outperforms Meta's Llama-3-8B-Instruct model, which was trained on over 10M pairs.
title Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2410.14251