@inproceedings{zhang-etal-2025-orthogonal,
title = "An Orthogonal High-Rank Adaptation for Large Language Models",
author = "Zhang, Xin and
Chen, Guang-Ze and
Li, Shuzhen and
Liu, Zhulin and
Chen, C.L.Philip and
Zhang, Tong",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.emnlp-main.951/",
pages = "18826--18844",
ISBN = "979-8-89176-332-6",
abstract = "Low-rank adaptation (LoRA) efficiently adapts LLMs to downstream tasks by decomposing LLMs' weight update into trainable low-rank matrices for fine-tuning. However, the random low-rank matrices may introduce massive task-irrelevant information, while their recomposed form suffer from limited representation spaces under low-rank operations. Such dense and choked adaptation in LoRA impairs the adaptation performance of LLMs on downstream tasks. To address these challenges, this paper proposes OHoRA, an orthogonal high-rank adaptation for parameter-efficient fine-tuning on LLMs. According to the information bottleneck theory, OHoRA decomposes LLMs' pre-trained weight matrices into orthogonal basis vectors via QR decomposition and splits them into two low-redundancy high-rank components to suppress task-irrelevant information. It then performs dynamic rank-elevated recomposition through Kronecker product to generate expansive task-tailored representation spaces, enabling precise LLM adaptation and enhanced generalization. OHoRA effectively operationalizes the information bottleneck theory to decompose LLMs' weight matrices into low-redundancy high-rank components and recompose them in rank-elevated manner for more task-tailored representation spaces and precise LLM adaptation. Empirical evaluation shows OHoRA{'}s effectiveness by outperforming LoRA and its variants and achieving comparable performance to full fine-tuning with only 0.0371{\%} trainable parameters."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2025-orthogonal">
<titleInfo>
<title>An Orthogonal High-Rank Adaptation for Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guang-Ze</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuzhen</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhulin</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">C.L.Philip</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-332-6</identifier>
</relatedItem>
<abstract>Low-rank adaptation (LoRA) efficiently adapts LLMs to downstream tasks by decomposing LLMs’ weight update into trainable low-rank matrices for fine-tuning. However, the random low-rank matrices may introduce massive task-irrelevant information, while their recomposed form suffer from limited representation spaces under low-rank operations. Such dense and choked adaptation in LoRA impairs the adaptation performance of LLMs on downstream tasks. To address these challenges, this paper proposes OHoRA, an orthogonal high-rank adaptation for parameter-efficient fine-tuning on LLMs. According to the information bottleneck theory, OHoRA decomposes LLMs’ pre-trained weight matrices into orthogonal basis vectors via QR decomposition and splits them into two low-redundancy high-rank components to suppress task-irrelevant information. It then performs dynamic rank-elevated recomposition through Kronecker product to generate expansive task-tailored representation spaces, enabling precise LLM adaptation and enhanced generalization. OHoRA effectively operationalizes the information bottleneck theory to decompose LLMs’ weight matrices into low-redundancy high-rank components and recompose them in rank-elevated manner for more task-tailored representation spaces and precise LLM adaptation. Empirical evaluation shows OHoRA’s effectiveness by outperforming LoRA and its variants and achieving comparable performance to full fine-tuning with only 0.0371% trainable parameters.</abstract>
<identifier type="citekey">zhang-etal-2025-orthogonal</identifier>
<location>
<url>https://aclanthology.org/2025.emnlp-main.951/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>18826</start>
<end>18844</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T An Orthogonal High-Rank Adaptation for Large Language Models
%A Zhang, Xin
%A Chen, Guang-Ze
%A Li, Shuzhen
%A Liu, Zhulin
%A Chen, C.L.Philip
%A Zhang, Tong
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-332-6
%F zhang-etal-2025-orthogonal
%X Low-rank adaptation (LoRA) efficiently adapts LLMs to downstream tasks by decomposing LLMs’ weight update into trainable low-rank matrices for fine-tuning. However, the random low-rank matrices may introduce massive task-irrelevant information, while their recomposed form suffer from limited representation spaces under low-rank operations. Such dense and choked adaptation in LoRA impairs the adaptation performance of LLMs on downstream tasks. To address these challenges, this paper proposes OHoRA, an orthogonal high-rank adaptation for parameter-efficient fine-tuning on LLMs. According to the information bottleneck theory, OHoRA decomposes LLMs’ pre-trained weight matrices into orthogonal basis vectors via QR decomposition and splits them into two low-redundancy high-rank components to suppress task-irrelevant information. It then performs dynamic rank-elevated recomposition through Kronecker product to generate expansive task-tailored representation spaces, enabling precise LLM adaptation and enhanced generalization. OHoRA effectively operationalizes the information bottleneck theory to decompose LLMs’ weight matrices into low-redundancy high-rank components and recompose them in rank-elevated manner for more task-tailored representation spaces and precise LLM adaptation. Empirical evaluation shows OHoRA’s effectiveness by outperforming LoRA and its variants and achieving comparable performance to full fine-tuning with only 0.0371% trainable parameters.
%U https://aclanthology.org/2025.emnlp-main.951/
%P 18826-18844
Markdown (Informal)
[An Orthogonal High-Rank Adaptation for Large Language Models](https://aclanthology.org/2025.emnlp-main.951/) (Zhang et al., EMNLP 2025)
ACL
- Xin Zhang, Guang-Ze Chen, Shuzhen Li, Zhulin Liu, C.L.Philip Chen, and Tong Zhang. 2025. An Orthogonal High-Rank Adaptation for Large Language Models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 18826–18844, Suzhou, China. Association for Computational Linguistics.