@inproceedings{agrawal-tandon-2026-lost,
title = "Lost in Embedding: Recovering Information Loss in Dense Retrieval",
author = "Agrawal, Prayas and
Tandon, Niket",
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.2307/",
doi = "10.18653/2026.findings-emnlp.2307",
pages = "49498--49517",
ISBN = "979-8-89176-462-0",
abstract = "Standard dense retrievers optimize to bring relevant query document pairs closer, but do not preserve which task-critical information (e.g., entities, dates, concepts, or logical relationships) survives compression into embeddings. We observe that supplementing queries with such information at inference time improves retrieval, motivating the question of how to train information-preserving dense retrievers. We propose a new training strategy that augments contrastive learning with a reconstruction loss to provide supervision over embedding content. Our method, RecovIR, is a decoder-only LLM retriever that appends learned latent tokens and uses their representations for both cosine retrieval and autoregressive reconstruction. Across multiple architectures, RecovIR consistently improves nDCG@10 by 1{--}2 points on MTEB-English retrieval benchmarks over standard baselines. We find that retrieval failures often coincide with reconstruction errors, and show theoretically that when targets are relevance-sufficient and recoverable, reconstruction lower-bounds the relevance information retained in embeddings."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="agrawal-tandon-2026-lost">
<titleInfo>
<title>Lost in Embedding: Recovering Information Loss in Dense Retrieval</title>
</titleInfo>
<name type="personal">
<namePart type="given">Prayas</namePart>
<namePart type="family">Agrawal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Niket</namePart>
<namePart type="family">Tandon</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>Standard dense retrievers optimize to bring relevant query document pairs closer, but do not preserve which task-critical information (e.g., entities, dates, concepts, or logical relationships) survives compression into embeddings. We observe that supplementing queries with such information at inference time improves retrieval, motivating the question of how to train information-preserving dense retrievers. We propose a new training strategy that augments contrastive learning with a reconstruction loss to provide supervision over embedding content. Our method, RecovIR, is a decoder-only LLM retriever that appends learned latent tokens and uses their representations for both cosine retrieval and autoregressive reconstruction. Across multiple architectures, RecovIR consistently improves nDCG@10 by 1–2 points on MTEB-English retrieval benchmarks over standard baselines. We find that retrieval failures often coincide with reconstruction errors, and show theoretically that when targets are relevance-sufficient and recoverable, reconstruction lower-bounds the relevance information retained in embeddings.</abstract>
<identifier type="citekey">agrawal-tandon-2026-lost</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2307</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2307/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>49498</start>
<end>49517</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Lost in Embedding: Recovering Information Loss in Dense Retrieval
%A Agrawal, Prayas
%A Tandon, Niket
%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 agrawal-tandon-2026-lost
%X Standard dense retrievers optimize to bring relevant query document pairs closer, but do not preserve which task-critical information (e.g., entities, dates, concepts, or logical relationships) survives compression into embeddings. We observe that supplementing queries with such information at inference time improves retrieval, motivating the question of how to train information-preserving dense retrievers. We propose a new training strategy that augments contrastive learning with a reconstruction loss to provide supervision over embedding content. Our method, RecovIR, is a decoder-only LLM retriever that appends learned latent tokens and uses their representations for both cosine retrieval and autoregressive reconstruction. Across multiple architectures, RecovIR consistently improves nDCG@10 by 1–2 points on MTEB-English retrieval benchmarks over standard baselines. We find that retrieval failures often coincide with reconstruction errors, and show theoretically that when targets are relevance-sufficient and recoverable, reconstruction lower-bounds the relevance information retained in embeddings.
%R 10.18653/2026.findings-emnlp.2307
%U https://aclanthology.org/2026.findings-emnlp.2307/
%U https://doi.org/10.18653/2026.findings-emnlp.2307
%P 49498-49517
Markdown (Informal)
[Lost in Embedding: Recovering Information Loss in Dense Retrieval](https://aclanthology.org/2026.findings-emnlp.2307/) (Agrawal & Tandon, Findings 2026)
ACL