A Model with No Head and Many Thoughts

Nikita Koriagin; Yaroslav Aksenov; George Bredis; Gleb Gerasimov; Nikita Balagansky; Daniil Gavrilov

A Model with No Head and Many Thoughts

Nikita Koriagin, Yaroslav Aksenov, George Bredis, Gleb Gerasimov, Nikita Balagansky, Daniil Gavrilov

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

Large language models decode by projecting hidden states through a large vocabulary head at every step. This operation is computationally costly and forces all reasoning to be expressed in discrete tokens. We introduce Soft Latent Thinking, a method that replaces the LM head during reasoning with a lightweight projector, enabling autoregressive rollout in embedding space where reasoning steps remain continuous rather than tokenized. Experiments on DeepSeek-Qwen-1.5B and LLaMA-3.2-3B show that Soft Latent Thinking consistently improves pass@k across all k while reducing per-step compute during chain-of-thought. Our method achieves the highest pass@32 among all soft-thinking approaches, demonstrating that effective reasoning can be carried out in continuous space without discrete token generation.

Anthology ID:: 2026.findings-emnlp.430
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:: 9058–9067
Language:
URL:: https://aclanthology.org/2026.findings-emnlp.430/
DOI:
Bibkey:
Cite (ACL):: Nikita Koriagin, Yaroslav Aksenov, George Bredis, Gleb Gerasimov, Nikita Balagansky, and Daniil Gavrilov. 2026. A Model with No Head and Many Thoughts. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 9058–9067, Budapest, Hungary. Association for Computational Linguistics.
Cite (Informal):: A Model with No Head and Many Thoughts (Koriagin et al., Findings 2026)
Copy Citation:
PDF:: https://aclanthology.org/2026.findings-emnlp.430.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{koriagin-etal-2026-model,
    title = "A Model with No Head and Many Thoughts",
    author = "Koriagin, Nikita  and
      Aksenov, Yaroslav  and
      Bredis, George  and
      Gerasimov, Gleb  and
      Balagansky, Nikita  and
      Gavrilov, Daniil",
    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.430/",
    pages = "9058--9067",
    ISBN = "979-8-89176-462-0",
    abstract = "Large language models decode by projecting hidden states through a large vocabulary head at every step. This operation is computationally costly and forces all reasoning to be expressed in discrete tokens. We introduce Soft Latent Thinking, a method that replaces the LM head during reasoning with a lightweight projector, enabling autoregressive rollout in embedding space where reasoning steps remain continuous rather than tokenized. Experiments on DeepSeek-Qwen-1.5B and LLaMA-3.2-3B show that Soft Latent Thinking consistently improves pass@k across all k while reducing per-step compute during chain-of-thought. Our method achieves the highest pass@32 among all soft-thinking approaches, demonstrating that effective reasoning can be carried out in continuous space without discrete token generation."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="koriagin-etal-2026-model">
    <titleInfo>
        <title>A Model with No Head and Many Thoughts</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Nikita</namePart>
        <namePart type="family">Koriagin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yaroslav</namePart>
        <namePart type="family">Aksenov</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">George</namePart>
        <namePart type="family">Bredis</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Gleb</namePart>
        <namePart type="family">Gerasimov</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nikita</namePart>
        <namePart type="family">Balagansky</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Daniil</namePart>
        <namePart type="family">Gavrilov</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>Large language models decode by projecting hidden states through a large vocabulary head at every step. This operation is computationally costly and forces all reasoning to be expressed in discrete tokens. We introduce Soft Latent Thinking, a method that replaces the LM head during reasoning with a lightweight projector, enabling autoregressive rollout in embedding space where reasoning steps remain continuous rather than tokenized. Experiments on DeepSeek-Qwen-1.5B and LLaMA-3.2-3B show that Soft Latent Thinking consistently improves pass@k across all k while reducing per-step compute during chain-of-thought. Our method achieves the highest pass@32 among all soft-thinking approaches, demonstrating that effective reasoning can be carried out in continuous space without discrete token generation.</abstract>
    <identifier type="citekey">koriagin-etal-2026-model</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.430/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>9058</start>
            <end>9067</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T A Model with No Head and Many Thoughts
%A Koriagin, Nikita
%A Aksenov, Yaroslav
%A Bredis, George
%A Gerasimov, Gleb
%A Balagansky, Nikita
%A Gavrilov, Daniil
%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 koriagin-etal-2026-model
%X Large language models decode by projecting hidden states through a large vocabulary head at every step. This operation is computationally costly and forces all reasoning to be expressed in discrete tokens. We introduce Soft Latent Thinking, a method that replaces the LM head during reasoning with a lightweight projector, enabling autoregressive rollout in embedding space where reasoning steps remain continuous rather than tokenized. Experiments on DeepSeek-Qwen-1.5B and LLaMA-3.2-3B show that Soft Latent Thinking consistently improves pass@k across all k while reducing per-step compute during chain-of-thought. Our method achieves the highest pass@32 among all soft-thinking approaches, demonstrating that effective reasoning can be carried out in continuous space without discrete token generation.
%U https://aclanthology.org/2026.findings-emnlp.430/
%P 9058-9067

Download as File

Markdown (Informal)

[A Model with No Head and Many Thoughts](https://aclanthology.org/2026.findings-emnlp.430/) (Koriagin et al., Findings 2026)

A Model with No Head and Many Thoughts (Koriagin et al., Findings 2026)

ACL

Nikita Koriagin, Yaroslav Aksenov, George Bredis, Gleb Gerasimov, Nikita Balagansky, and Daniil Gavrilov. 2026. A Model with No Head and Many Thoughts. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 9058–9067, Budapest, Hungary. Association for Computational Linguistics.