@inproceedings{michel-etal-2026-reading,
title = "Reading Between the Turns: User Repair as a Context-Dependent Signal of Breakdown and Recovery in Multi-Turn {LLM} Assistants",
author = "Michel, Leah and
Pacheco, Nuno Miguel Martins and
Ulan uulu, Choro and
Kulyabin, Mikhail and
Joosten, Jan and
Petridis, Filippos and
Hedderich, Michael A. and
Plank, Barbara",
editor = "Sousa-Silva, Rui and
Lopes Cardoso, Henrique and
Koponen, Maarit and
Pareja-Lora, Antonio",
booktitle = "Proceedings of the 3rd Workshop on Language Understanding in the Human-Machine Era ({LUHME} 2026)",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.luhme-1.3/",
pages = "38--53",
ISBN = "979-8-89176-471-2",
abstract = "In multi-turn interactions with deployed LLM assistants, users rarely provide explicit feedback, making it difficult to assess whether the interaction succeeded or failed. We therefore investigate user repair: follow-up user actions such as rephrasing, correcting, clarifying, or requesting more information. Such actions may signal breakdown recovery, but they can also reflect ordinary task progression, making them difficult to interpret in isolation. Using a human-annotated dataset of industrial multi-turn conversations, we examine when repair is informative about assistant-side deficiencies and user dissatisfaction. We find that repair occurs more often after deficient responses, but it is also common after non-deficient responses, making it a weak standalone signal of failure. Its diagnostic value becomes clearer when it is interpreted jointly with assistant response quality, repair type, and dialogue context. Repetition/rephrasing is most strongly associated with negative feedback on deficient responses, whereas requests for more information are more common after non-deficient responses, suggesting a role in collaborative task progression. We further show that repair is not always local: while most repairs target the immediately preceding turn, a substantial minority target earlier dialogue context. These findings position user repair as a context-dependent interactional signal for monitoring, evaluating, and diagnosing deployed assistants."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="michel-etal-2026-reading">
<titleInfo>
<title>Reading Between the Turns: User Repair as a Context-Dependent Signal of Breakdown and Recovery in Multi-Turn LLM Assistants</title>
</titleInfo>
<name type="personal">
<namePart type="given">Leah</namePart>
<namePart type="family">Michel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nuno</namePart>
<namePart type="given">Miguel</namePart>
<namePart type="given">Martins</namePart>
<namePart type="family">Pacheco</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Choro</namePart>
<namePart type="family">Ulan uulu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mikhail</namePart>
<namePart type="family">Kulyabin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jan</namePart>
<namePart type="family">Joosten</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Filippos</namePart>
<namePart type="family">Petridis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="given">A</namePart>
<namePart type="family">Hedderich</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="family">Plank</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 3rd Workshop on Language Understanding in the Human-Machine Era (LUHME 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Sousa-Silva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henrique</namePart>
<namePart type="family">Lopes Cardoso</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maarit</namePart>
<namePart type="family">Koponen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Pareja-Lora</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">979-8-89176-471-2</identifier>
</relatedItem>
<abstract>In multi-turn interactions with deployed LLM assistants, users rarely provide explicit feedback, making it difficult to assess whether the interaction succeeded or failed. We therefore investigate user repair: follow-up user actions such as rephrasing, correcting, clarifying, or requesting more information. Such actions may signal breakdown recovery, but they can also reflect ordinary task progression, making them difficult to interpret in isolation. Using a human-annotated dataset of industrial multi-turn conversations, we examine when repair is informative about assistant-side deficiencies and user dissatisfaction. We find that repair occurs more often after deficient responses, but it is also common after non-deficient responses, making it a weak standalone signal of failure. Its diagnostic value becomes clearer when it is interpreted jointly with assistant response quality, repair type, and dialogue context. Repetition/rephrasing is most strongly associated with negative feedback on deficient responses, whereas requests for more information are more common after non-deficient responses, suggesting a role in collaborative task progression. We further show that repair is not always local: while most repairs target the immediately preceding turn, a substantial minority target earlier dialogue context. These findings position user repair as a context-dependent interactional signal for monitoring, evaluating, and diagnosing deployed assistants.</abstract>
<identifier type="citekey">michel-etal-2026-reading</identifier>
<location>
<url>https://aclanthology.org/2026.luhme-1.3/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>38</start>
<end>53</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Reading Between the Turns: User Repair as a Context-Dependent Signal of Breakdown and Recovery in Multi-Turn LLM Assistants
%A Michel, Leah
%A Pacheco, Nuno Miguel Martins
%A Ulan uulu, Choro
%A Kulyabin, Mikhail
%A Joosten, Jan
%A Petridis, Filippos
%A Hedderich, Michael A.
%A Plank, Barbara
%Y Sousa-Silva, Rui
%Y Lopes Cardoso, Henrique
%Y Koponen, Maarit
%Y Pareja-Lora, Antonio
%S Proceedings of the 3rd Workshop on Language Understanding in the Human-Machine Era (LUHME 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-471-2
%F michel-etal-2026-reading
%X In multi-turn interactions with deployed LLM assistants, users rarely provide explicit feedback, making it difficult to assess whether the interaction succeeded or failed. We therefore investigate user repair: follow-up user actions such as rephrasing, correcting, clarifying, or requesting more information. Such actions may signal breakdown recovery, but they can also reflect ordinary task progression, making them difficult to interpret in isolation. Using a human-annotated dataset of industrial multi-turn conversations, we examine when repair is informative about assistant-side deficiencies and user dissatisfaction. We find that repair occurs more often after deficient responses, but it is also common after non-deficient responses, making it a weak standalone signal of failure. Its diagnostic value becomes clearer when it is interpreted jointly with assistant response quality, repair type, and dialogue context. Repetition/rephrasing is most strongly associated with negative feedback on deficient responses, whereas requests for more information are more common after non-deficient responses, suggesting a role in collaborative task progression. We further show that repair is not always local: while most repairs target the immediately preceding turn, a substantial minority target earlier dialogue context. These findings position user repair as a context-dependent interactional signal for monitoring, evaluating, and diagnosing deployed assistants.
%U https://aclanthology.org/2026.luhme-1.3/
%P 38-53
Markdown (Informal)
[Reading Between the Turns: User Repair as a Context-Dependent Signal of Breakdown and Recovery in Multi-Turn LLM Assistants](https://aclanthology.org/2026.luhme-1.3/) (Michel et al., LUHME 2026)
ACL
- Leah Michel, Nuno Miguel Martins Pacheco, Choro Ulan uulu, Mikhail Kulyabin, Jan Joosten, Filippos Petridis, Michael A. Hedderich, and Barbara Plank. 2026. Reading Between the Turns: User Repair as a Context-Dependent Signal of Breakdown and Recovery in Multi-Turn LLM Assistants. In Proceedings of the 3rd Workshop on Language Understanding in the Human-Machine Era (LUHME 2026), pages 38–53, Budapest, Hungary. Association for Computational Linguistics.