@inproceedings{yu-etal-2026-improving,
title = "Improving Low-resource Grammatical Error Correction Via Automatic Linguistic Neurons Probing and Optimizing",
author = "Yu, Zihao and
Li, Ying and
Yu, Zhengtao and
Gao, Shengxiang and
Mao, Cunli and
Huang, Yuxin and
Lai, Hua and
Wang, Wenjun",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.909/",
pages = "19241--19252",
ISBN = "979-8-89176-462-0",
abstract = "Large Language Models (LLMs) exhibit strong language understanding and instruction-following abilities, making them promising backbones for grammatical error correction (GEC). However, prompt-only correction for low-resource languages still suffers from the over-correction challenge, since LLMs cannot fully understand their internal grammar. To address this issue, we propose an automatic linguistic neuron probing and optimizing approach to improve the low-resource GEC with only a few neuron activations. Concretely, we first utilize multiple low-resource GEC corpora to observe the changes in neurons, thus obtaining a linguistically specific probe matrix to ensure whether the neuron is activated. Then, we construct an extra weight matrix to determine the activation intensity of each neuron. Finally, we combine the probe matrix and the weight matrix to obtain an activation matrix, which is injected into selected model layers during inference to guide the frozen LLM toward more controllable GEC. Experiments on four benchmark datasets show that our approach consistently improves GEC performance, indicating that lightweight sparse activation rescaling is extremely useful for low-resource grammatical knowledge understanding. Detailed analysis demonstrates that our approach can further provide interpretable neuron-level control without large-scale parameter fine-tuning."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yu-etal-2026-improving">
<titleInfo>
<title>Improving Low-resource Grammatical Error Correction Via Automatic Linguistic Neurons Probing and Optimizing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zihao</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ying</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhengtao</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shengxiang</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cunli</namePart>
<namePart type="family">Mao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxin</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hua</namePart>
<namePart type="family">Lai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenjun</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) exhibit strong language understanding and instruction-following abilities, making them promising backbones for grammatical error correction (GEC). However, prompt-only correction for low-resource languages still suffers from the over-correction challenge, since LLMs cannot fully understand their internal grammar. To address this issue, we propose an automatic linguistic neuron probing and optimizing approach to improve the low-resource GEC with only a few neuron activations. Concretely, we first utilize multiple low-resource GEC corpora to observe the changes in neurons, thus obtaining a linguistically specific probe matrix to ensure whether the neuron is activated. Then, we construct an extra weight matrix to determine the activation intensity of each neuron. Finally, we combine the probe matrix and the weight matrix to obtain an activation matrix, which is injected into selected model layers during inference to guide the frozen LLM toward more controllable GEC. Experiments on four benchmark datasets show that our approach consistently improves GEC performance, indicating that lightweight sparse activation rescaling is extremely useful for low-resource grammatical knowledge understanding. Detailed analysis demonstrates that our approach can further provide interpretable neuron-level control without large-scale parameter fine-tuning.</abstract>
<identifier type="citekey">yu-etal-2026-improving</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.909/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>19241</start>
<end>19252</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Improving Low-resource Grammatical Error Correction Via Automatic Linguistic Neurons Probing and Optimizing
%A Yu, Zihao
%A Li, Ying
%A Yu, Zhengtao
%A Gao, Shengxiang
%A Mao, Cunli
%A Huang, Yuxin
%A Lai, Hua
%A Wang, Wenjun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F yu-etal-2026-improving
%X Large Language Models (LLMs) exhibit strong language understanding and instruction-following abilities, making them promising backbones for grammatical error correction (GEC). However, prompt-only correction for low-resource languages still suffers from the over-correction challenge, since LLMs cannot fully understand their internal grammar. To address this issue, we propose an automatic linguistic neuron probing and optimizing approach to improve the low-resource GEC with only a few neuron activations. Concretely, we first utilize multiple low-resource GEC corpora to observe the changes in neurons, thus obtaining a linguistically specific probe matrix to ensure whether the neuron is activated. Then, we construct an extra weight matrix to determine the activation intensity of each neuron. Finally, we combine the probe matrix and the weight matrix to obtain an activation matrix, which is injected into selected model layers during inference to guide the frozen LLM toward more controllable GEC. Experiments on four benchmark datasets show that our approach consistently improves GEC performance, indicating that lightweight sparse activation rescaling is extremely useful for low-resource grammatical knowledge understanding. Detailed analysis demonstrates that our approach can further provide interpretable neuron-level control without large-scale parameter fine-tuning.
%U https://aclanthology.org/2026.findings-emnlp.909/
%P 19241-19252
Markdown (Informal)
[Improving Low-resource Grammatical Error Correction Via Automatic Linguistic Neurons Probing and Optimizing](https://aclanthology.org/2026.findings-emnlp.909/) (Yu et al., Findings 2026)
ACL
- Zihao Yu, Ying Li, Zhengtao Yu, Shengxiang Gao, Cunli Mao, Yuxin Huang, Hua Lai, and Wenjun Wang. 2026. Improving Low-resource Grammatical Error Correction Via Automatic Linguistic Neurons Probing and Optimizing. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 19241–19252, Budapest, Hungary. Association for Computational Linguistics.