ReviewInstruct: A Review-Driven Multi-Turn Conversations Generation Method for Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Jiangxu, Wang, Cong, Su, TianHuang, Yang, Jun, Lin, Haozhi, Zhang, Chao, Peng, Ming, Shi, Kai, Yang, SongPan, Pan, BinQing, Li, ZiXian, Yang, Ni, Yang, ZhenYu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908433161125888
author Wu, Jiangxu
Wang, Cong
Su, TianHuang
Yang, Jun
Lin, Haozhi
Zhang, Chao
Peng, Ming
Shi, Kai
Yang, SongPan
Pan, BinQing
Li, ZiXian
Yang, Ni
Yang, ZhenYu
author_facet Wu, Jiangxu
Wang, Cong
Su, TianHuang
Yang, Jun
Lin, Haozhi
Zhang, Chao
Peng, Ming
Shi, Kai
Yang, SongPan
Pan, BinQing
Li, ZiXian
Yang, Ni
Yang, ZhenYu
contents The effectiveness of large language models (LLMs) in conversational AI is hindered by their reliance on single-turn supervised fine-tuning (SFT) data, which limits contextual coherence in multi-turn dialogues. Existing methods for generating multi-turn dialogue data struggle to ensure both diversity and quality in instructions. To address this, we propose Review-Instruct, a novel framework that synthesizes multi-turn conversations through an iterative "Ask-Respond-Review" process involving three agent roles: a Candidate, multiple Reviewers, and a Chairman. The framework iteratively refines instructions by incorporating Reviewer feedback, enhancing dialogue diversity and difficulty. We construct a multi-turn dataset using the Alpaca dataset and fine-tune the LLaMA2-13B model. Evaluations on MT-Bench, MMLU-Pro, and Auto-Arena demonstrate significant improvements, achieving absolute gains of 2.9\% on MMLU-Pro and 2\% on MT-Bench compared to prior state-of-the-art models based on LLaMA2-13B. Ablation studies confirm the critical role of the Review stage and the use of multiple Reviewers in boosting instruction diversity and difficulty. Our work highlights the potential of review-driven, multi-agent frameworks for generating high-quality conversational data at scale.
format Preprint
id arxiv_https___arxiv_org_abs_2505_11010
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ReviewInstruct: A Review-Driven Multi-Turn Conversations Generation Method for Large Language Models
Wu, Jiangxu
Wang, Cong
Su, TianHuang
Yang, Jun
Lin, Haozhi
Zhang, Chao
Peng, Ming
Shi, Kai
Yang, SongPan
Pan, BinQing
Li, ZiXian
Yang, Ni
Yang, ZhenYu
Computation and Language
Artificial Intelligence
The effectiveness of large language models (LLMs) in conversational AI is hindered by their reliance on single-turn supervised fine-tuning (SFT) data, which limits contextual coherence in multi-turn dialogues. Existing methods for generating multi-turn dialogue data struggle to ensure both diversity and quality in instructions. To address this, we propose Review-Instruct, a novel framework that synthesizes multi-turn conversations through an iterative "Ask-Respond-Review" process involving three agent roles: a Candidate, multiple Reviewers, and a Chairman. The framework iteratively refines instructions by incorporating Reviewer feedback, enhancing dialogue diversity and difficulty. We construct a multi-turn dataset using the Alpaca dataset and fine-tune the LLaMA2-13B model. Evaluations on MT-Bench, MMLU-Pro, and Auto-Arena demonstrate significant improvements, achieving absolute gains of 2.9\% on MMLU-Pro and 2\% on MT-Bench compared to prior state-of-the-art models based on LLaMA2-13B. Ablation studies confirm the critical role of the Review stage and the use of multiple Reviewers in boosting instruction diversity and difficulty. Our work highlights the potential of review-driven, multi-agent frameworks for generating high-quality conversational data at scale.
title ReviewInstruct: A Review-Driven Multi-Turn Conversations Generation Method for Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.11010