<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="aswal-etal-2026-observable">
    <titleInfo>
        <title>Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Darpan</namePart>
        <namePart type="family">Aswal</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Thomas</namePart>
        <namePart type="given">Palmeira</namePart>
        <namePart type="family">Ferraz</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yongxin</namePart>
        <namePart type="family">Zhou</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Maxime</namePart>
        <namePart type="family">Peyrard</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>Latent reasoning models (LRMs) replace explicit chain-of-thought with continuous thoughts. Recent work treats observable latent-state patterns, such as BFS-like frontiers and decodable arithmetic computation, as evidence for internal reasoning mechanisms. Evaluating two LRMs (Coconut and CODI) against controls lacking the proposed recurrence or curriculum, we find these patterns also appear in the controls and do not consistently affect behavior. Causal interventions reveal that latent-thought utilization is not binary but graded, scaling with a thought’s causal effect on model behavior. Geometric analyses reveal this effect concentrates in low-dimensional subspaces whose step-wise geometry grows more structured as their behavioral influence increases. Latent thoughts should therefore be treated as hidden computation, not hidden explanation: decodability, attention, or static structure alone cannot establish LRM reasoning mechanisms; that requires matched controls and causal tests.</abstract>
    <identifier type="citekey">aswal-etal-2026-observable</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.83/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>1898</start>
            <end>1938</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models
%A Aswal, Darpan
%A Ferraz, Thomas Palmeira
%A Zhou, Yongxin
%A Peyrard, Maxime
%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 aswal-etal-2026-observable
%X Latent reasoning models (LRMs) replace explicit chain-of-thought with continuous thoughts. Recent work treats observable latent-state patterns, such as BFS-like frontiers and decodable arithmetic computation, as evidence for internal reasoning mechanisms. Evaluating two LRMs (Coconut and CODI) against controls lacking the proposed recurrence or curriculum, we find these patterns also appear in the controls and do not consistently affect behavior. Causal interventions reveal that latent-thought utilization is not binary but graded, scaling with a thought’s causal effect on model behavior. Geometric analyses reveal this effect concentrates in low-dimensional subspaces whose step-wise geometry grows more structured as their behavioral influence increases. Latent thoughts should therefore be treated as hidden computation, not hidden explanation: decodability, attention, or static structure alone cannot establish LRM reasoning mechanisms; that requires matched controls and causal tests.
%U https://aclanthology.org/2026.emnlp-main.83/
%P 1898-1938

Markdown (Informal)

[Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models](https://aclanthology.org/2026.emnlp-main.83/) (Aswal et al., EMNLP 2026)

• Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models (Aswal et al., EMNLP 2026)

ACL

• Darpan Aswal, Thomas Palmeira Ferraz, Yongxin Zhou, and Maxime Peyrard. 2026. Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 1898–1938, Budapest, Hungary. Association for Computational Linguistics.