Triplet-Block Diffusion RWKV

Ke Lin; Yiyang Luo; Zhaolong Su; Yunya Song; Anyi Rao

Triplet-Block Diffusion RWKV

Ke Lin, Yiyang Luo, Zhaolong Su, Yunya Song, Anyi Rao

Correct Metadata for

Use this form to create a GitHub issue with structured data describing the correction. You will need a GitHub account. Once you create that issue, the correction will be reviewed by a staff member.

⚠️ Mobile Users: Submitting this form to create a new issue will only work with github.com, not the GitHub Mobile app.

Important: The Anthology treat PDFs as authoritative. Please use this form only to correct data that is out of line with the PDF. See our corrections guidelines if you need to change the PDF.

Title Adjust the title. Retain tags such as <fixed-case>.

Authors Adjust author names and order to match the PDF.

Abstract Correct abstract if needed. Retain XML formatting tags such as <tex-math>. You may use ... for bold, ... for italic, ... for underline, <sc>...</sc> for small-caps, <tt>...<tt> for typewriter text, <url>...</url> for URLs, <a href=...> for hyperlinks, and <par/> for paragraph breaks.

Verification against PDF Ensure that the new title/authors match the snapshot below. (If there is no snapshot or it is too small, consult the PDF.)

Authors concatenated from the text boxes above:

ALL author names match the snapshot above—including middle initials, hyphens, and accents.

Abstract

Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost, resulting in expensive inference and an inability to revise previous tokens during text generation. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent and underdeveloped. In this work, we propose B3D-RWKV, a diffusion-based RWKV variant that integrates the model’s 𝒪(L) inference efficiency with parallel, bidirectional discrete-diffusion through a triplet-block layout method. B3D-RWKV-7.2B reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of 1.6× speedup. Code is available at https://github.com/leonardodalinky/B3D-RWKV.

Anthology ID:: 2026.findings-emnlp.940
Volume:: Findings of the Association for Computational Linguistics: EMNLP 2026
Month:: October
Year:: 2026
Address:: Budapest, Hungary
Editors:: Hung-Yi Lee, Sunipa Dev, Jessy Li, Desmond Elliott
Venue:: Findings
SIG:: SIGDAT
Publisher:: Association for Computational Linguistics
Note:
Pages:: 19816–19829
Language:
URL:: https://aclanthology.org/2026.findings-emnlp.940/
DOI:
Bibkey:
Cite (ACL):: Ke Lin, Yiyang Luo, Zhaolong Su, Yunya Song, and Anyi Rao. 2026. Triplet-Block Diffusion RWKV. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 19816–19829, Budapest, Hungary. Association for Computational Linguistics.
Cite (Informal):: Triplet-Block Diffusion RWKV (Lin et al., Findings 2026)
Copy Citation:
PDF:: https://aclanthology.org/2026.findings-emnlp.940.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{lin-etal-2026-triplet,
    title = "Triplet-Block Diffusion {RWKV}",
    author = "Lin, Ke  and
      Luo, Yiyang  and
      Su, Zhaolong  and
      Song, Yunya  and
      Rao, Anyi",
    editor = "Lee, Hung-Yi  and
      Dev, Sunipa  and
      Li, Jessy  and
      Elliott, Desmond",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
    month = oct,
    year = "2026",
    address = "Budapest, Hungary",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.findings-emnlp.940/",
    pages = "19816--19829",
    ISBN = "979-8-89176-462-0",
    abstract = "Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost, resulting in expensive inference and an inability to revise previous tokens during text generation. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent and underdeveloped. In this work, we propose B3D-RWKV, a diffusion-based RWKV variant that integrates the model{'}s $\mathcal{O}(L)$ inference efficiency with parallel, bidirectional discrete-diffusion through a triplet-block layout method. B3D-RWKV-7.2B reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of $1.6\times$ speedup. Code is available at https://github.com/leonardodalinky/B3D-RWKV."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lin-etal-2026-triplet">
    <titleInfo>
        <title>Triplet-Block Diffusion RWKV</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Ke</namePart>
        <namePart type="family">Lin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yiyang</namePart>
        <namePart type="family">Luo</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhaolong</namePart>
        <namePart type="family">Su</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yunya</namePart>
        <namePart type="family">Song</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Anyi</namePart>
        <namePart type="family">Rao</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost, resulting in expensive inference and an inability to revise previous tokens during text generation. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent and underdeveloped. In this work, we propose B3D-RWKV, a diffusion-based RWKV variant that integrates the model’s \mathcalO(L) inference efficiency with parallel, bidirectional discrete-diffusion through a triplet-block layout method. B3D-RWKV-7.2B reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of 1.6\times speedup. Code is available at https://github.com/leonardodalinky/B3D-RWKV.</abstract>
    <identifier type="citekey">lin-etal-2026-triplet</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.940/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>19816</start>
            <end>19829</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T Triplet-Block Diffusion RWKV
%A Lin, Ke
%A Luo, Yiyang
%A Su, Zhaolong
%A Song, Yunya
%A Rao, Anyi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F lin-etal-2026-triplet
%X Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost, resulting in expensive inference and an inability to revise previous tokens during text generation. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent and underdeveloped. In this work, we propose B3D-RWKV, a diffusion-based RWKV variant that integrates the model’s \mathcalO(L) inference efficiency with parallel, bidirectional discrete-diffusion through a triplet-block layout method. B3D-RWKV-7.2B reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of 1.6\times speedup. Code is available at https://github.com/leonardodalinky/B3D-RWKV.
%U https://aclanthology.org/2026.findings-emnlp.940/
%P 19816-19829

Download as File

Markdown (Informal)

[Triplet-Block Diffusion RWKV](https://aclanthology.org/2026.findings-emnlp.940/) (Lin et al., Findings 2026)

Triplet-Block Diffusion RWKV (Lin et al., Findings 2026)

ACL

Ke Lin, Yiyang Luo, Zhaolong Su, Yunya Song, and Anyi Rao. 2026. Triplet-Block Diffusion RWKV. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 19816–19829, Budapest, Hungary. Association for Computational Linguistics.