ENTP: Enhancing Low-Quality SFT Data via Neural-Symbolic Text Purge-Mix

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Zile, Li, Ling, Di, Na, Pang, Jinlong, Zhou, Yao, Cheng, Hao, Han, Bo, Wei, Jiaheng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909870884651008
author Yang, Zile
Li, Ling
Di, Na
Pang, Jinlong
Zhou, Yao
Cheng, Hao
Han, Bo
Wei, Jiaheng
author_facet Yang, Zile
Li, Ling
Di, Na
Pang, Jinlong
Zhou, Yao
Cheng, Hao
Han, Bo
Wei, Jiaheng
contents Supervised Fine-Tuning (SFT) adapts pre-trained Large Language Models (LLMs) to domain-specific instructions by training on a carefully curated subset of high-quality instruction-response pairs, typically drawn from a larger dataset that often contains many low-quality or noisy samples. However, existing quality-first paradigms often overlook valuable signals in discarded low-quality data and rely on imperfect quality filters. We introduce ENTP (Enhancing low-quality SFT data via Neural-symbolic Text Purge-Mix), a framework that revitalizes low-quality corpora through symbolic purification and neural reconstruction. The symbolic module identifies and prunes noisy samples based on statistical priors, while the neural component synthesizes enriched instruction-response pairs by leveraging latent representations and model knowledge. This neural-symbolic synergy enhances data informativeness and diversity. Experiments show that ENTP-augmented datasets, constructed exclusively from low-quality data, outperform 13 established data-selection baselines across five instruction-following benchmarks, and even surpass fine-tuning on the full original dataset (approximately 300K examples). Our results highlight the untapped potential of low-quality data and underscore the importance of intelligent purification and synthesis for efficient instruction alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2510_23160
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ENTP: Enhancing Low-Quality SFT Data via Neural-Symbolic Text Purge-Mix
Yang, Zile
Li, Ling
Di, Na
Pang, Jinlong
Zhou, Yao
Cheng, Hao
Han, Bo
Wei, Jiaheng
Computation and Language
Supervised Fine-Tuning (SFT) adapts pre-trained Large Language Models (LLMs) to domain-specific instructions by training on a carefully curated subset of high-quality instruction-response pairs, typically drawn from a larger dataset that often contains many low-quality or noisy samples. However, existing quality-first paradigms often overlook valuable signals in discarded low-quality data and rely on imperfect quality filters. We introduce ENTP (Enhancing low-quality SFT data via Neural-symbolic Text Purge-Mix), a framework that revitalizes low-quality corpora through symbolic purification and neural reconstruction. The symbolic module identifies and prunes noisy samples based on statistical priors, while the neural component synthesizes enriched instruction-response pairs by leveraging latent representations and model knowledge. This neural-symbolic synergy enhances data informativeness and diversity. Experiments show that ENTP-augmented datasets, constructed exclusively from low-quality data, outperform 13 established data-selection baselines across five instruction-following benchmarks, and even surpass fine-tuning on the full original dataset (approximately 300K examples). Our results highlight the untapped potential of low-quality data and underscore the importance of intelligent purification and synthesis for efficient instruction alignment.
title ENTP: Enhancing Low-Quality SFT Data via Neural-Symbolic Text Purge-Mix
topic Computation and Language
url https://arxiv.org/abs/2510.23160