<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lin-etal-2026-spectr">
    <titleInfo>
        <title>SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yijun</namePart>
        <namePart type="family">Lin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jin-Hao</namePart>
        <namePart type="family">Sheng</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qingyue</namePart>
        <namePart type="family">Cai</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Feng</namePart>
        <namePart type="family">Zhou</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>Autoregressive language models suffer from high inference latency due to their sequential decoding nature. Speculative decoding (SD) mitigates this by employing a lightweight draft model to propose candidate tokens, which are selectively verified by a larger target model. While existing methods either adopt multi-draft strategies to increase acceptance rates or block verification techniques to jointly verify multiple tokens, they remain limited by treating these improvements in isolation. In this work, we propose SpecTr-GBV, a novel SD method that unifies multi-draft and greedy block verification (GBV) into a single framework. By formulating the verification step as an optimal transport problem over draft and target token blocks, SpecTr-GBV improves both theoretical efficiency and empirical performance. We theoretically prove that SpecTr-GBV achieves the optimal expected acceptance length within the class of conditionally independent couplings (CIC) induced by our multi-draft verification design, and show that this optimal value increases with the number of drafts. Empirically, we evaluate SpecTr-GBV across five datasets and four baselines. Our method achieves superior speedup and significantly higher block efficiency while preserving output quality. In addition, we perform comprehensive ablation studies to evaluate the impact of various hyperparameters in the model. The code is publicly available at https://github.com/junemily/SpecTr-GBV.git.</abstract>
    <identifier type="citekey">lin-etal-2026-spectr</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1340/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>30562</start>
            <end>30578</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding
%A Lin, Yijun
%A Sheng, Jin-Hao
%A Cai, Qingyue
%A Zhou, Feng
%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 lin-etal-2026-spectr
%X Autoregressive language models suffer from high inference latency due to their sequential decoding nature. Speculative decoding (SD) mitigates this by employing a lightweight draft model to propose candidate tokens, which are selectively verified by a larger target model. While existing methods either adopt multi-draft strategies to increase acceptance rates or block verification techniques to jointly verify multiple tokens, they remain limited by treating these improvements in isolation. In this work, we propose SpecTr-GBV, a novel SD method that unifies multi-draft and greedy block verification (GBV) into a single framework. By formulating the verification step as an optimal transport problem over draft and target token blocks, SpecTr-GBV improves both theoretical efficiency and empirical performance. We theoretically prove that SpecTr-GBV achieves the optimal expected acceptance length within the class of conditionally independent couplings (CIC) induced by our multi-draft verification design, and show that this optimal value increases with the number of drafts. Empirically, we evaluate SpecTr-GBV across five datasets and four baselines. Our method achieves superior speedup and significantly higher block efficiency while preserving output quality. In addition, we perform comprehensive ablation studies to evaluate the impact of various hyperparameters in the model. The code is publicly available at https://github.com/junemily/SpecTr-GBV.git.
%U https://aclanthology.org/2026.emnlp-main.1340/
%P 30562-30578

Markdown (Informal)

[SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding](https://aclanthology.org/2026.emnlp-main.1340/) (Lin et al., EMNLP 2026)

• SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding (Lin et al., EMNLP 2026)

ACL

• Yijun Lin, Jin-Hao Sheng, Qingyue Cai, and Feng Zhou. 2026. SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 30562–30578, Budapest, Hungary. Association for Computational Linguistics.