<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2025-opt">
    <titleInfo>
        <title>OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structure</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jikai</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yi</namePart>
        <namePart type="family">Su</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Juntao</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qingrong</namePart>
        <namePart type="family">Xia</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zi</namePart>
        <namePart type="family">Ye</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xinyu</namePart>
        <namePart type="family">Duan</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhefeng</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Min</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <genre authority="bibutilsgt">journal article</genre>
    <relatedItem type="host">
        <titleInfo>
            <title>Transactions of the Association for Computational Linguistics</title>
        </titleInfo>
        <originInfo>
            <issuance>continuing</issuance>
            <publisher>MIT Press</publisher>
            <place>
                <placeTerm type="text">Cambridge, MA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">periodical</genre>
        <genre authority="bibutilsgt">academic journal</genre>
    </relatedItem>
    <abstract>Autoregressive language models demonstrate excellent performance in various scenarios. However, the inference efficiency is limited by its one-step-one-word generation mode, which has become a pressing problem recently as the models become increasingly larger. Speculative decoding employs a “draft and then verify” mechanism to allow multiple tokens to be generated in one step, realizing lossless acceleration. Existing methods mainly adopt fixed heuristic draft structures, which do not adapt to different situations to maximize the acceptance length during verification. To alleviate this dilemma, we propose OPT-Tree, an algorithm to construct adaptive and scalable draft trees, which can be applied to any autoregressive draft model. It searches the optimal tree structure that maximizes the mathematical expectation of the acceptance length in each decoding step. Experimental results reveal that OPT-Tree outperforms the existing draft structures and achieves a speed-up ratio of up to 3.2 compared with autoregressive decoding. If the draft model is powerful enough and the node budget is sufficient, it can generate more than ten tokens in a single step. Our code is available at https://github.com/Jikai0Wang/OPT-Tree.</abstract>
    <identifier type="citekey">wang-etal-2025-opt</identifier>
    <identifier type="doi">10.1162/tacl_a_00735</identifier>
    <location>
        <url>https://aclanthology.org/2025.tacl-1.8/</url>
    </location>
    <part>
        <date>2025</date>
        <detail type="volume"><number>13</number></detail>
        <extent unit="page">
            <start>188</start>
            <end>199</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Journal Article
%T OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structure
%A Wang, Jikai
%A Su, Yi
%A Li, Juntao
%A Xia, Qingrong
%A Ye, Zi
%A Duan, Xinyu
%A Wang, Zhefeng
%A Zhang, Min
%J Transactions of the Association for Computational Linguistics
%D 2025
%V 13
%I MIT Press
%C Cambridge, MA
%F wang-etal-2025-opt
%X Autoregressive language models demonstrate excellent performance in various scenarios. However, the inference efficiency is limited by its one-step-one-word generation mode, which has become a pressing problem recently as the models become increasingly larger. Speculative decoding employs a “draft and then verify” mechanism to allow multiple tokens to be generated in one step, realizing lossless acceleration. Existing methods mainly adopt fixed heuristic draft structures, which do not adapt to different situations to maximize the acceptance length during verification. To alleviate this dilemma, we propose OPT-Tree, an algorithm to construct adaptive and scalable draft trees, which can be applied to any autoregressive draft model. It searches the optimal tree structure that maximizes the mathematical expectation of the acceptance length in each decoding step. Experimental results reveal that OPT-Tree outperforms the existing draft structures and achieves a speed-up ratio of up to 3.2 compared with autoregressive decoding. If the draft model is powerful enough and the node budget is sufficient, it can generate more than ten tokens in a single step. Our code is available at https://github.com/Jikai0Wang/OPT-Tree.
%R 10.1162/tacl_a_00735
%U https://aclanthology.org/2025.tacl-1.8/
%U https://doi.org/10.1162/tacl_a_00735
%P 188-199

Markdown (Informal)

[OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structure](https://aclanthology.org/2025.tacl-1.8/) (Wang et al., TACL 2025)

• OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structure (Wang et al., TACL 2025)

ACL

• Jikai Wang, Yi Su, Juntao Li, Qingrong Xia, Zi Ye, Xinyu Duan, Zhefeng Wang, and Min Zhang. 2025. OPT-Tree: Speculative Decoding with Adaptive Draft Tree Structure. Transactions of the Association for Computational Linguistics, 13:188–199.