@inproceedings{zhang-zhu-2026-anchors,
title = "{ANCHORS}: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control",
author = "Zhang, Yuxuan and
Zhu, Jiaqi",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1215/",
pages = "27755--27771",
ISBN = "979-8-89176-459-0",
abstract = "While Parameter-Efficient Fine-Tuning (PEFT) reduces the cost of adapting Large Language Models (LLMs), current methods like LoRA suffer from catastrophic forgetting due to unregulated updates on high-importance parameters. To address this stability-plasticity dilemma, we introduce ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control, a novel framework that ensures uninterrupted knowledge retention during aggressive adaptation. ANCHORS integrates CABR Decomposition to preserve informative subspaces and S-MagNorm, a bio-inspired normalization mechanism that utilizes Sigmoid-guided magnitude control to dynamically dampen updates on critical weights. Extensive experiments on reasoning and QA benchmarks demonstrate that ANCHORS consistently outperforms LoRA, DoRA, and CUR-LoRA. It achieves performance gains of up to 4.05{\%} in MCQ and 2.61{\%} in QA tasks while retaining 61.12{\%} of pre-trained knowledge{---}surpassing Experience Replay and EWC baselines in continuous learning scenarios."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-zhu-2026-anchors">
<titleInfo>
<title>ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaqi</namePart>
<namePart type="family">Zhu</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>While Parameter-Efficient Fine-Tuning (PEFT) reduces the cost of adapting Large Language Models (LLMs), current methods like LoRA suffer from catastrophic forgetting due to unregulated updates on high-importance parameters. To address this stability-plasticity dilemma, we introduce ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control, a novel framework that ensures uninterrupted knowledge retention during aggressive adaptation. ANCHORS integrates CABR Decomposition to preserve informative subspaces and S-MagNorm, a bio-inspired normalization mechanism that utilizes Sigmoid-guided magnitude control to dynamically dampen updates on critical weights. Extensive experiments on reasoning and QA benchmarks demonstrate that ANCHORS consistently outperforms LoRA, DoRA, and CUR-LoRA. It achieves performance gains of up to 4.05% in MCQ and 2.61% in QA tasks while retaining 61.12% of pre-trained knowledge—surpassing Experience Replay and EWC baselines in continuous learning scenarios.</abstract>
<identifier type="citekey">zhang-zhu-2026-anchors</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1215/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>27755</start>
<end>27771</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control
%A Zhang, Yuxuan
%A Zhu, Jiaqi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F zhang-zhu-2026-anchors
%X While Parameter-Efficient Fine-Tuning (PEFT) reduces the cost of adapting Large Language Models (LLMs), current methods like LoRA suffer from catastrophic forgetting due to unregulated updates on high-importance parameters. To address this stability-plasticity dilemma, we introduce ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control, a novel framework that ensures uninterrupted knowledge retention during aggressive adaptation. ANCHORS integrates CABR Decomposition to preserve informative subspaces and S-MagNorm, a bio-inspired normalization mechanism that utilizes Sigmoid-guided magnitude control to dynamically dampen updates on critical weights. Extensive experiments on reasoning and QA benchmarks demonstrate that ANCHORS consistently outperforms LoRA, DoRA, and CUR-LoRA. It achieves performance gains of up to 4.05% in MCQ and 2.61% in QA tasks while retaining 61.12% of pre-trained knowledge—surpassing Experience Replay and EWC baselines in continuous learning scenarios.
%U https://aclanthology.org/2026.emnlp-main.1215/
%P 27755-27771
Markdown (Informal)
[ANCHORS: Adaptive Normalization for Continual High-Order Retention via Sigmoid-Guided Magnitude Control](https://aclanthology.org/2026.emnlp-main.1215/) (Zhang & Zhu, EMNLP 2026)
ACL