@inproceedings{zhou-etal-2026-queries,
title = "When Queries Crowd: Diagnosing and Improving Spread-Out Regularization for Text Embeddings",
author = "Zhou, Zeqi and
Zeng, Qingcheng and
Yang, Rui and
Liu, Jiarui and
Qi, Heli and
Yu, Puxuan and
Xuan, Weihao and
Yokoya, Naoto",
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.1585/",
pages = "36051--36070",
ISBN = "979-8-89176-459-0",
abstract = "Text embedding models are central to information retrieval and deep research tasks, and their behaviors are strongly shaped by the training objective. While recent systems have largely converged on InfoNCE-style loss functions, there have been continued attempts to design additional loss functions beyond this core objective. One such component is spread-out loss, a batch-level penalty on high pairwise similarity that encourages embeddings to occupy the representation space more evenly. However, in existing systems, this design is usually combined with many other design choices and has not been sufficiently ablated. To isolate its effect, we use spread-out loss as a controlled lens for studying geometric regularization in text embedding training. Across 13 retrieval benchmarks, spread-out improves fine-tuned retrievers over contrastive baselines across architectures and model sizes. We then diagnose where these gains arise and find that they concentrate on short queries, where spread-out reduces query-side directional crowding. Based on this diagnosis, we propose QUARTO, a Query-length Adaptive spread-out loss that uses batch composition to scale the full regularizer more strongly for short-query-dominated batches and less strongly for longer-query-dominated batches. QUARTO further improves retrieval performance over fixed-weight spread-out. Overall, our findings point to the potential of embedding-space regularization and suggest that text-embedding training should explore a broader design space of objectives beyond InfoNCE-style contrastive learning. Project page: https://github.com/JoeZhou-encoding/Spreadout{\_}QUARTO"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhou-etal-2026-queries">
<titleInfo>
<title>When Queries Crowd: Diagnosing and Improving Spread-Out Regularization for Text Embeddings</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zeqi</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingcheng</namePart>
<namePart type="family">Zeng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiarui</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heli</namePart>
<namePart type="family">Qi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Puxuan</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weihao</namePart>
<namePart type="family">Xuan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoto</namePart>
<namePart type="family">Yokoya</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>Text embedding models are central to information retrieval and deep research tasks, and their behaviors are strongly shaped by the training objective. While recent systems have largely converged on InfoNCE-style loss functions, there have been continued attempts to design additional loss functions beyond this core objective. One such component is spread-out loss, a batch-level penalty on high pairwise similarity that encourages embeddings to occupy the representation space more evenly. However, in existing systems, this design is usually combined with many other design choices and has not been sufficiently ablated. To isolate its effect, we use spread-out loss as a controlled lens for studying geometric regularization in text embedding training. Across 13 retrieval benchmarks, spread-out improves fine-tuned retrievers over contrastive baselines across architectures and model sizes. We then diagnose where these gains arise and find that they concentrate on short queries, where spread-out reduces query-side directional crowding. Based on this diagnosis, we propose QUARTO, a Query-length Adaptive spread-out loss that uses batch composition to scale the full regularizer more strongly for short-query-dominated batches and less strongly for longer-query-dominated batches. QUARTO further improves retrieval performance over fixed-weight spread-out. Overall, our findings point to the potential of embedding-space regularization and suggest that text-embedding training should explore a broader design space of objectives beyond InfoNCE-style contrastive learning. Project page: https://github.com/JoeZhou-encoding/Spreadout_QUARTO</abstract>
<identifier type="citekey">zhou-etal-2026-queries</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1585/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>36051</start>
<end>36070</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T When Queries Crowd: Diagnosing and Improving Spread-Out Regularization for Text Embeddings
%A Zhou, Zeqi
%A Zeng, Qingcheng
%A Yang, Rui
%A Liu, Jiarui
%A Qi, Heli
%A Yu, Puxuan
%A Xuan, Weihao
%A Yokoya, Naoto
%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 zhou-etal-2026-queries
%X Text embedding models are central to information retrieval and deep research tasks, and their behaviors are strongly shaped by the training objective. While recent systems have largely converged on InfoNCE-style loss functions, there have been continued attempts to design additional loss functions beyond this core objective. One such component is spread-out loss, a batch-level penalty on high pairwise similarity that encourages embeddings to occupy the representation space more evenly. However, in existing systems, this design is usually combined with many other design choices and has not been sufficiently ablated. To isolate its effect, we use spread-out loss as a controlled lens for studying geometric regularization in text embedding training. Across 13 retrieval benchmarks, spread-out improves fine-tuned retrievers over contrastive baselines across architectures and model sizes. We then diagnose where these gains arise and find that they concentrate on short queries, where spread-out reduces query-side directional crowding. Based on this diagnosis, we propose QUARTO, a Query-length Adaptive spread-out loss that uses batch composition to scale the full regularizer more strongly for short-query-dominated batches and less strongly for longer-query-dominated batches. QUARTO further improves retrieval performance over fixed-weight spread-out. Overall, our findings point to the potential of embedding-space regularization and suggest that text-embedding training should explore a broader design space of objectives beyond InfoNCE-style contrastive learning. Project page: https://github.com/JoeZhou-encoding/Spreadout_QUARTO
%U https://aclanthology.org/2026.emnlp-main.1585/
%P 36051-36070
Markdown (Informal)
[When Queries Crowd: Diagnosing and Improving Spread-Out Regularization for Text Embeddings](https://aclanthology.org/2026.emnlp-main.1585/) (Zhou et al., EMNLP 2026)
ACL
- Zeqi Zhou, Qingcheng Zeng, Rui Yang, Jiarui Liu, Heli Qi, Puxuan Yu, Weihao Xuan, and Naoto Yokoya. 2026. When Queries Crowd: Diagnosing and Improving Spread-Out Regularization for Text Embeddings. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 36051–36070, Budapest, Hungary. Association for Computational Linguistics.