@inproceedings{nguyen-etal-2026-tracing,
title = "Tracing Error Propagation in Reasoning Language Models",
author = "Nguyen, Dung Huu and
Pham, Quan and
Nguyen, Dat Quoc",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1559/",
pages = "35544--35559",
ISBN = "979-8-89176-459-0",
abstract = "Reasoning language models (RLMs) increasingly tackle complex problems through long reasoning traces, but identifying when and where failures arise remains difficult. In this work, we study whether reasoning errors are signaled in model representations before they manifest as failed traces. We first show that step-level errors are locally predictable within a short window of neighboring steps, suggesting that models encode early warning signals. However, trace-level failures often result not from isolated errors, but from persistent ones originating earlier in the trace. To study this, we distinguish *recoverable* from *unrecoverable* traces via counterfactual resampling. We find that unrecoverable traces induce final-step representations that are clearly separable from those of recoverable traces. Moreover, the persistent error can be retrospectively decoded from the final-step representation, indicating that error signals propagate and accumulate through the reasoning trace."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nguyen-etal-2026-tracing">
<titleInfo>
<title>Tracing Error Propagation in Reasoning Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dung</namePart>
<namePart type="given">Huu</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Quan</namePart>
<namePart type="family">Pham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dat</namePart>
<namePart type="given">Quoc</namePart>
<namePart type="family">Nguyen</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>Reasoning language models (RLMs) increasingly tackle complex problems through long reasoning traces, but identifying when and where failures arise remains difficult. In this work, we study whether reasoning errors are signaled in model representations before they manifest as failed traces. We first show that step-level errors are locally predictable within a short window of neighboring steps, suggesting that models encode early warning signals. However, trace-level failures often result not from isolated errors, but from persistent ones originating earlier in the trace. To study this, we distinguish *recoverable* from *unrecoverable* traces via counterfactual resampling. We find that unrecoverable traces induce final-step representations that are clearly separable from those of recoverable traces. Moreover, the persistent error can be retrospectively decoded from the final-step representation, indicating that error signals propagate and accumulate through the reasoning trace.</abstract>
<identifier type="citekey">nguyen-etal-2026-tracing</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1559/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>35544</start>
<end>35559</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Tracing Error Propagation in Reasoning Language Models
%A Nguyen, Dung Huu
%A Pham, Quan
%A Nguyen, Dat Quoc
%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 nguyen-etal-2026-tracing
%X Reasoning language models (RLMs) increasingly tackle complex problems through long reasoning traces, but identifying when and where failures arise remains difficult. In this work, we study whether reasoning errors are signaled in model representations before they manifest as failed traces. We first show that step-level errors are locally predictable within a short window of neighboring steps, suggesting that models encode early warning signals. However, trace-level failures often result not from isolated errors, but from persistent ones originating earlier in the trace. To study this, we distinguish *recoverable* from *unrecoverable* traces via counterfactual resampling. We find that unrecoverable traces induce final-step representations that are clearly separable from those of recoverable traces. Moreover, the persistent error can be retrospectively decoded from the final-step representation, indicating that error signals propagate and accumulate through the reasoning trace.
%U https://aclanthology.org/2026.emnlp-main.1559/
%P 35544-35559
Markdown (Informal)
[Tracing Error Propagation in Reasoning Language Models](https://aclanthology.org/2026.emnlp-main.1559/) (Nguyen et al., EMNLP 2026)
ACL
- Dung Huu Nguyen, Quan Pham, and Dat Quoc Nguyen. 2026. Tracing Error Propagation in Reasoning Language Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 35544–35559, Budapest, Hungary. Association for Computational Linguistics.