<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-hcspec">
    <titleInfo>
        <title>HCSpec: Two-Tier Horizontal Cascade Speculative Decoding for High-Efficiency Large Language Model Inference</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yizhou</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Siming</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hao</namePart>
        <namePart type="family">Ye</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Erhu</namePart>
        <namePart type="family">Feng</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Maria</namePart>
            <namePart type="family">Liakata</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Viviane</namePart>
            <namePart type="given">P</namePart>
            <namePart type="family">Moreira</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jiajun</namePart>
            <namePart type="family">Zhang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">David</namePart>
            <namePart type="family">Jurgens</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, California, United States</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-390-6</identifier>
    </relatedItem>
    <abstract>Speculative decoding accelerates large language model (LLM) inference by using a draft model to propose token candidates for parallel verification by the target model. However, current state-of-the-art self-distilled draft models adopt a homogeneous architecture across all drafting positions, failing to account for a critical empirical observation: the expected utility of drafting decays rapidly after the initial positions. To exploit this imbalance, we propose Two-tier Horizontal Cascade Speculative Decoding (HCSpec), a novel framework that organizes heterogeneous, position-specialized draft modules into a horizontal cascade. The first tier employs a dual-layer, dual-path transformer that enhances early-step fidelity by decoupling token-logit prediction from recurrent feature propagation, while the second tier adopts a lightweight single-layer transformer that deliberately trades marginal accuracy for improved efficiency at later drafting steps. Extensive experiments on Qwen series models and Llama3.1-8B-Instruct, across multiple tasks and diverse inference configurations, demonstrate that HCSpec consistently outperforms the previous state-of-the-art (EAGLE-3). It delivers 15–30% higher end-to-end speedup over EAGLE-3 and achieves up to 3.72x acceleration over vanilla autoregressive decoding. Our code is provided in the supplementary materials.</abstract>
    <identifier type="citekey">zhang-etal-2026-hcspec</identifier>
    <location>
        <url>https://aclanthology.org/2026.acl-long.353/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>7773</start>
            <end>7783</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T HCSpec: Two-Tier Horizontal Cascade Speculative Decoding for High-Efficiency Large Language Model Inference
%A Zhang, Yizhou
%A Chen, Siming
%A Ye, Hao
%A Feng, Erhu
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F zhang-etal-2026-hcspec
%X Speculative decoding accelerates large language model (LLM) inference by using a draft model to propose token candidates for parallel verification by the target model. However, current state-of-the-art self-distilled draft models adopt a homogeneous architecture across all drafting positions, failing to account for a critical empirical observation: the expected utility of drafting decays rapidly after the initial positions. To exploit this imbalance, we propose Two-tier Horizontal Cascade Speculative Decoding (HCSpec), a novel framework that organizes heterogeneous, position-specialized draft modules into a horizontal cascade. The first tier employs a dual-layer, dual-path transformer that enhances early-step fidelity by decoupling token-logit prediction from recurrent feature propagation, while the second tier adopts a lightweight single-layer transformer that deliberately trades marginal accuracy for improved efficiency at later drafting steps. Extensive experiments on Qwen series models and Llama3.1-8B-Instruct, across multiple tasks and diverse inference configurations, demonstrate that HCSpec consistently outperforms the previous state-of-the-art (EAGLE-3). It delivers 15–30% higher end-to-end speedup over EAGLE-3 and achieves up to 3.72x acceleration over vanilla autoregressive decoding. Our code is provided in the supplementary materials.
%U https://aclanthology.org/2026.acl-long.353/
%P 7773-7783

Markdown (Informal)

[HCSpec: Two-Tier Horizontal Cascade Speculative Decoding for High-Efficiency Large Language Model Inference](https://aclanthology.org/2026.acl-long.353/) (Zhang et al., ACL 2026)

• HCSpec: Two-Tier Horizontal Cascade Speculative Decoding for High-Efficiency Large Language Model Inference (Zhang et al., ACL 2026)

ACL

• Yizhou Zhang, Siming Chen, Hao Ye, and Erhu Feng. 2026. HCSpec: Two-Tier Horizontal Cascade Speculative Decoding for High-Efficiency Large Language Model Inference. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7773–7783, San Diego, California, United States. Association for Computational Linguistics.