@inproceedings{vu-etal-2026-sugar,
title = "{SUGAR}: Spectral and Geometry-Aware Alignment for Matryoshka Representation Distillation",
author = "Vu, Hai An and
Huy, Phung Gia and
Truong, Minh-Phuc and
Van, Linh Ngo and
Nguyen, Thanh Hong and
Do, Thanh-Toan and
Le, Trung",
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.228/",
pages = "5257--5271",
ISBN = "979-8-89176-459-0",
abstract = "Matryoshka Representation Learning (MRL) produces nested embeddings whose prefixes remain semantically meaningful, allowing representations to be truncated at inference time and enabling a single model to operate across multiple dimensional budgets. This paradigm has been adopted in several modern embedding models such as Qwen3-Embedding and OpenAI text-embedding-3. However, distilling MRL teachers into compact students remains largely unexplored, as conventional knowledge distillation methods focus on aligning single-scale embeddings and do not account for the multi-scale structure of Matryoshka representations. We introduce Matryoshka Representation Distillation and propose SUGAR, a new framework that preserves the structural properties of MRL embeddings during knowledge distillation. SUGAR aligns point-wise semantic fidelity, spectral capacity via effective-rank matching, and global subspace geometry through Grassmannian alignment. Our extensive experiments show that SUGAR produces compact students that retain strong performance across multiple embedding dimensions while preserving deployment flexibility."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vu-etal-2026-sugar">
<titleInfo>
<title>SUGAR: Spectral and Geometry-Aware Alignment for Matryoshka Representation Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hai</namePart>
<namePart type="given">An</namePart>
<namePart type="family">Vu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Phung</namePart>
<namePart type="given">Gia</namePart>
<namePart type="family">Huy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minh-Phuc</namePart>
<namePart type="family">Truong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linh</namePart>
<namePart type="given">Ngo</namePart>
<namePart type="family">Van</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thanh</namePart>
<namePart type="given">Hong</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thanh-Toan</namePart>
<namePart type="family">Do</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Trung</namePart>
<namePart type="family">Le</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>Matryoshka Representation Learning (MRL) produces nested embeddings whose prefixes remain semantically meaningful, allowing representations to be truncated at inference time and enabling a single model to operate across multiple dimensional budgets. This paradigm has been adopted in several modern embedding models such as Qwen3-Embedding and OpenAI text-embedding-3. However, distilling MRL teachers into compact students remains largely unexplored, as conventional knowledge distillation methods focus on aligning single-scale embeddings and do not account for the multi-scale structure of Matryoshka representations. We introduce Matryoshka Representation Distillation and propose SUGAR, a new framework that preserves the structural properties of MRL embeddings during knowledge distillation. SUGAR aligns point-wise semantic fidelity, spectral capacity via effective-rank matching, and global subspace geometry through Grassmannian alignment. Our extensive experiments show that SUGAR produces compact students that retain strong performance across multiple embedding dimensions while preserving deployment flexibility.</abstract>
<identifier type="citekey">vu-etal-2026-sugar</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.228/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>5257</start>
<end>5271</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SUGAR: Spectral and Geometry-Aware Alignment for Matryoshka Representation Distillation
%A Vu, Hai An
%A Huy, Phung Gia
%A Truong, Minh-Phuc
%A Van, Linh Ngo
%A Nguyen, Thanh Hong
%A Do, Thanh-Toan
%A Le, Trung
%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 vu-etal-2026-sugar
%X Matryoshka Representation Learning (MRL) produces nested embeddings whose prefixes remain semantically meaningful, allowing representations to be truncated at inference time and enabling a single model to operate across multiple dimensional budgets. This paradigm has been adopted in several modern embedding models such as Qwen3-Embedding and OpenAI text-embedding-3. However, distilling MRL teachers into compact students remains largely unexplored, as conventional knowledge distillation methods focus on aligning single-scale embeddings and do not account for the multi-scale structure of Matryoshka representations. We introduce Matryoshka Representation Distillation and propose SUGAR, a new framework that preserves the structural properties of MRL embeddings during knowledge distillation. SUGAR aligns point-wise semantic fidelity, spectral capacity via effective-rank matching, and global subspace geometry through Grassmannian alignment. Our extensive experiments show that SUGAR produces compact students that retain strong performance across multiple embedding dimensions while preserving deployment flexibility.
%U https://aclanthology.org/2026.emnlp-main.228/
%P 5257-5271
Markdown (Informal)
[SUGAR: Spectral and Geometry-Aware Alignment for Matryoshka Representation Distillation](https://aclanthology.org/2026.emnlp-main.228/) (Vu et al., EMNLP 2026)
ACL