<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-learning-retrieve">
    <titleInfo>
        <title>Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yibo</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nikki</namePart>
        <namePart type="given">Lijing</namePart>
        <namePart type="family">Kuang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Philip</namePart>
        <namePart type="given">S</namePart>
        <namePart type="family">Yu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhewei</namePart>
        <namePart type="family">Yao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yuxiong</namePart>
        <namePart type="family">He</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>Interactive text-to-SQL agents solve database tasks through multi-turn interactions involving schema exploration, query execution, feedback interpretation, and decision revision. Long-term memory helps agents reuse past experiences, but existing retrieval methods remain limited. Static methods use fixed similarity heuristics, whereas dynamic methods often learn from sparse terminal outcomes and retrieve at a single decision horizon. This is insufficient when memory usefulness changes across interaction stages, since memories useful for initial planning may differ from those needed for local, state-conditioned execution. We propose MERIT, a dynamic multi-horizon memory retrieval framework. MERIT maintains episode-level memory for global strategic guidance and turn-level memory for local decision support. Both levels use learned retrieval policies optimized with reinforcement learning. To train turn-level retrieval despite limited intermediate supervision, MERIT uses a lightweight Process Reward Model to provide dense proxy rewards for local memory selection. On BIRD-Interact, MERIT achieves higher success rates and fewer interaction turns than no-memory, static-retrieval, and dynamic-retrieval baselines. Transfer results on Spider2-Snow further show that the learned policies transfer without degradation, staying above all memory baselines without benchmark-specific tuning. These results suggest that multi-horizon retrieval improves experience reuse in interactive text-to-SQL agents. Our code is available at https://github.com/YiboWANG214/MERIT.</abstract>
    <identifier type="citekey">wang-etal-2026-learning-retrieve</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1986/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>45174</start>
            <end>45197</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents
%A Wang, Yibo
%A Kuang, Nikki Lijing
%A Yu, Philip S.
%A Yao, Zhewei
%A He, Yuxiong
%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 wang-etal-2026-learning-retrieve
%X Interactive text-to-SQL agents solve database tasks through multi-turn interactions involving schema exploration, query execution, feedback interpretation, and decision revision. Long-term memory helps agents reuse past experiences, but existing retrieval methods remain limited. Static methods use fixed similarity heuristics, whereas dynamic methods often learn from sparse terminal outcomes and retrieve at a single decision horizon. This is insufficient when memory usefulness changes across interaction stages, since memories useful for initial planning may differ from those needed for local, state-conditioned execution. We propose MERIT, a dynamic multi-horizon memory retrieval framework. MERIT maintains episode-level memory for global strategic guidance and turn-level memory for local decision support. Both levels use learned retrieval policies optimized with reinforcement learning. To train turn-level retrieval despite limited intermediate supervision, MERIT uses a lightweight Process Reward Model to provide dense proxy rewards for local memory selection. On BIRD-Interact, MERIT achieves higher success rates and fewer interaction turns than no-memory, static-retrieval, and dynamic-retrieval baselines. Transfer results on Spider2-Snow further show that the learned policies transfer without degradation, staying above all memory baselines without benchmark-specific tuning. These results suggest that multi-horizon retrieval improves experience reuse in interactive text-to-SQL agents. Our code is available at https://github.com/YiboWANG214/MERIT.
%U https://aclanthology.org/2026.emnlp-main.1986/
%P 45174-45197

Markdown (Informal)

[Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents](https://aclanthology.org/2026.emnlp-main.1986/) (Wang et al., EMNLP 2026)

• Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents (Wang et al., EMNLP 2026)

ACL

• Yibo Wang, Nikki Lijing Kuang, Philip S. Yu, Zhewei Yao, and Yuxiong He. 2026. Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 45174–45197, Budapest, Hungary. Association for Computational Linguistics.