Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Zijun, Tu, Haoqin, Zhou, Weidong, Zhou, Yiyang, Zhou, Xiaohuan, Zhang, Bingni, Feng, Weiguo, Wang, Taifeng, Xie, Cihang, Liu, Fengze
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2604.15706
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915941896421376
author Wang, Zijun
Tu, Haoqin
Zhou, Weidong
Zhou, Yiyang
Zhou, Xiaohuan
Zhang, Bingni
Feng, Weiguo
Wang, Taifeng
Xie, Cihang
Liu, Fengze
author_facet Wang, Zijun
Tu, Haoqin
Zhou, Weidong
Zhou, Yiyang
Zhou, Xiaohuan
Zhang, Bingni
Feng, Weiguo
Wang, Taifeng
Xie, Cihang
Liu, Fengze
contents Everyday tasks come with a target, and pretraining models around this target is what turns them into experts. In this paper, we study target-oriented language model (LM) pretraining by introducing Neuron-Activated Graph Ranking (NAG-based Ranking), a training-free and interpretable framework for target pretraining data selection. Rather than using black-box representations, our approach directly characterizes each target input by a sparse set of high-impact neurons in any off-the-shelf LLMs. Concretely, we quantify neuron impact and select the most influential neurons across layers into a compact Neuron-Activated Graph (NAG), and rank candidate data by NAG similarity to target examples. We conduct experiments across six benchmarks, where our NAG-based Ranking improves target-oriented pretraining by 4.9% on average over random sampling, and also outperforms state-of-the-art baselines by 5.3% accuracy on HellaSwag. It also remains effective under a more applicable multi-target setting, where our best setup surpasses two baselines by 1.1% and 4.1%, respectively. Furthermore, we provide a comprehensive analysis on why and how our NAG works, e.g., deactivating NAG-selected neurons (only 0.12% of all) causes a 23.5% performance collapse, and restricting NAG to the final layer incurs a 4.1% average drop, indicating that NAG captures a sparse "functional backbone" for learning target features. We release the code at https://github.com/asillycat/NAG.
format Preprint
id arxiv_https___arxiv_org_abs_2604_15706
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Target-Oriented Pretraining Data Selection via Neuron-Activated Graph
Wang, Zijun
Tu, Haoqin
Zhou, Weidong
Zhou, Yiyang
Zhou, Xiaohuan
Zhang, Bingni
Feng, Weiguo
Wang, Taifeng
Xie, Cihang
Liu, Fengze
Computation and Language
Everyday tasks come with a target, and pretraining models around this target is what turns them into experts. In this paper, we study target-oriented language model (LM) pretraining by introducing Neuron-Activated Graph Ranking (NAG-based Ranking), a training-free and interpretable framework for target pretraining data selection. Rather than using black-box representations, our approach directly characterizes each target input by a sparse set of high-impact neurons in any off-the-shelf LLMs. Concretely, we quantify neuron impact and select the most influential neurons across layers into a compact Neuron-Activated Graph (NAG), and rank candidate data by NAG similarity to target examples. We conduct experiments across six benchmarks, where our NAG-based Ranking improves target-oriented pretraining by 4.9% on average over random sampling, and also outperforms state-of-the-art baselines by 5.3% accuracy on HellaSwag. It also remains effective under a more applicable multi-target setting, where our best setup surpasses two baselines by 1.1% and 4.1%, respectively. Furthermore, we provide a comprehensive analysis on why and how our NAG works, e.g., deactivating NAG-selected neurons (only 0.12% of all) causes a 23.5% performance collapse, and restricting NAG to the final layer incurs a 4.1% average drop, indicating that NAG captures a sparse "functional backbone" for learning target features. We release the code at https://github.com/asillycat/NAG.
title Target-Oriented Pretraining Data Selection via Neuron-Activated Graph
topic Computation and Language
url https://arxiv.org/abs/2604.15706