<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wu-etal-2026-instance">
    <titleInfo>
        <title>From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Haodong</namePart>
        <namePart type="family">Wu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiahao</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Lijie</namePart>
        <namePart type="family">Hu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yongqi</namePart>
        <namePart type="family">Zhang</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>Supervised fine-tuning (SFT) data selection is commonly formulated as instance ranking: score each example and retain a top-k subset. However, effective SFT training subsets are often produced through ordered curation recipes, where filtering, mixing, and deduplication operators jointly shape the final data distribution. We formulate this problem as fixed-pool data recipe search: given a raw instruction pool and a library of grounded operators, the goal is to discover an executable recipe that constructs a high-quality selected subset under a limited budget of full SFT evaluations, without generating, rewriting, or augmenting training samples. We introduce AutoSelection, a two-layer solver that decouples fixed-pool materialization based on cached task-, data-, and model-side signals from expensive full evaluation, using warmup probes, realized subset states, local recipe edits, Gaussian-process-assisted ranking, and stagnation-triggered reseeding. Experiments on a 90K instruction pool show that AutoSelection achieves the strongest in-distribution reasoning average across three base models, outperforming full-data training, random recipe search, random top-k, and single-operator selectors. Additional held-out reasoning, instruction-following, graph-reasoning, search-stability, structural-ablation, and 1.5B-to-7B transfer results further show that recipe structure matters beyond individual selection operators. Code is available at https://github.com/w253/AutoSelection.</abstract>
    <identifier type="citekey">wu-etal-2026-instance</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1877/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>42785</start>
            <end>42818</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning
%A Wu, Haodong
%A Zhang, Jiahao
%A Hu, Lijie
%A Zhang, Yongqi
%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 wu-etal-2026-instance
%X Supervised fine-tuning (SFT) data selection is commonly formulated as instance ranking: score each example and retain a top-k subset. However, effective SFT training subsets are often produced through ordered curation recipes, where filtering, mixing, and deduplication operators jointly shape the final data distribution. We formulate this problem as fixed-pool data recipe search: given a raw instruction pool and a library of grounded operators, the goal is to discover an executable recipe that constructs a high-quality selected subset under a limited budget of full SFT evaluations, without generating, rewriting, or augmenting training samples. We introduce AutoSelection, a two-layer solver that decouples fixed-pool materialization based on cached task-, data-, and model-side signals from expensive full evaluation, using warmup probes, realized subset states, local recipe edits, Gaussian-process-assisted ranking, and stagnation-triggered reseeding. Experiments on a 90K instruction pool show that AutoSelection achieves the strongest in-distribution reasoning average across three base models, outperforming full-data training, random recipe search, random top-k, and single-operator selectors. Additional held-out reasoning, instruction-following, graph-reasoning, search-stability, structural-ablation, and 1.5B-to-7B transfer results further show that recipe structure matters beyond individual selection operators. Code is available at https://github.com/w253/AutoSelection.
%U https://aclanthology.org/2026.emnlp-main.1877/
%P 42785-42818

Markdown (Informal)

[From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning](https://aclanthology.org/2026.emnlp-main.1877/) (Wu et al., EMNLP 2026)

• From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning (Wu et al., EMNLP 2026)

ACL

• Haodong Wu, Jiahao Zhang, Lijie Hu, and Yongqi Zhang. 2026. From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 42785–42818, Budapest, Hungary. Association for Computational Linguistics.