@inproceedings{doshi-etal-2026-influence,
title = "Influence Guided Sampling for Domain Adaptation of Text Retrievers",
author = "Doshi, Meet and
Kumar, Vishwajeet and
Li, Yulong and
Sen, Jaydeep",
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.1957/",
pages = "41730--41752",
ISBN = "979-8-89176-462-0",
abstract = "General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose **Inf-DDS**, a novel reinforcement learning{--}driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight w.r.t. GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being $1.5\text{—}4\times$ cheaper in GPU compute. Our sampling strategy achieves a $\mathbf{5.03}$ absolute NDCG@10 improvement while training the multilingual bge-m3 model and an absolute NDCG@10 improvement of $\mathbf{0.94}$ while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="doshi-etal-2026-influence">
<titleInfo>
<title>Influence Guided Sampling for Domain Adaptation of Text Retrievers</title>
</titleInfo>
<name type="personal">
<namePart type="given">Meet</namePart>
<namePart type="family">Doshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vishwajeet</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yulong</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jaydeep</namePart>
<namePart type="family">Sen</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>General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose **Inf-DDS**, a novel reinforcement learning–driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight w.r.t. GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being 1.5\text—4\times cheaper in GPU compute. Our sampling strategy achieves a \mathbf5.03 absolute NDCG@10 improvement while training the multilingual bge-m3 model and an absolute NDCG@10 improvement of \mathbf0.94 while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets.</abstract>
<identifier type="citekey">doshi-etal-2026-influence</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1957/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>41730</start>
<end>41752</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Influence Guided Sampling for Domain Adaptation of Text Retrievers
%A Doshi, Meet
%A Kumar, Vishwajeet
%A Li, Yulong
%A Sen, Jaydeep
%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 doshi-etal-2026-influence
%X General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose **Inf-DDS**, a novel reinforcement learning–driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight w.r.t. GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being 1.5\text—4\times cheaper in GPU compute. Our sampling strategy achieves a \mathbf5.03 absolute NDCG@10 improvement while training the multilingual bge-m3 model and an absolute NDCG@10 improvement of \mathbf0.94 while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets.
%U https://aclanthology.org/2026.findings-emnlp.1957/
%P 41730-41752
Markdown (Informal)
[Influence Guided Sampling for Domain Adaptation of Text Retrievers](https://aclanthology.org/2026.findings-emnlp.1957/) (Doshi et al., Findings 2026)
ACL