@inproceedings{sivakumaran-etal-2026-balancing,
title = "Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution",
author = "Sivakumaran, Nithin and
Yu, Shoubin and
Lee, Hyunji and
Zhang, Yue and
Payani, Ali and
Bansal, Mohit and
Stengel-Eskin, Elias",
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.2140/",
pages = "45711--45730",
ISBN = "979-8-89176-462-0",
abstract = "Chain-of-thought (CoT) reasoning sometimes fails to faithfully reflect the true computation of a large language model (LLM), hampering its utility in explaining how LLMs arrive at their answers. Moreover, optimizing for faithfulness and interpretability in reasoning often degrades task performance. To address this tradeoff and improve CoT faithfulness, we propose Reasoning Execution by Multiple Listeners (REMuL), a multi-party reinforcement learning approach. REMuL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful. A speaker model generates a reasoning trace, which is truncated and passed to a pool of listener models who ``execute'' the trace, continuing the trace to an answer. Speakers are rewarded for producing reasoning that is clear to listeners, with additional correctness regularization via masked supervised finetuning to counter the tradeoff between faithfulness and performance. On multiple reasoning benchmarks (BIG-Bench Extra Hard, MuSR, ZebraLogicBench, and FOLIO), REMuL consistently and substantially improves three measures of faithfulness {--} hint attribution, early answering area over the curve (AOC), and mistake injection AOC {--} while also improving accuracy. Our analysis finds that these gains are robust across training domains and are associated with shorter and more direct CoTs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sivakumaran-etal-2026-balancing">
<titleInfo>
<title>Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nithin</namePart>
<namePart type="family">Sivakumaran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shoubin</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hyunji</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ali</namePart>
<namePart type="family">Payani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elias</namePart>
<namePart type="family">Stengel-Eskin</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>Chain-of-thought (CoT) reasoning sometimes fails to faithfully reflect the true computation of a large language model (LLM), hampering its utility in explaining how LLMs arrive at their answers. Moreover, optimizing for faithfulness and interpretability in reasoning often degrades task performance. To address this tradeoff and improve CoT faithfulness, we propose Reasoning Execution by Multiple Listeners (REMuL), a multi-party reinforcement learning approach. REMuL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful. A speaker model generates a reasoning trace, which is truncated and passed to a pool of listener models who “execute” the trace, continuing the trace to an answer. Speakers are rewarded for producing reasoning that is clear to listeners, with additional correctness regularization via masked supervised finetuning to counter the tradeoff between faithfulness and performance. On multiple reasoning benchmarks (BIG-Bench Extra Hard, MuSR, ZebraLogicBench, and FOLIO), REMuL consistently and substantially improves three measures of faithfulness – hint attribution, early answering area over the curve (AOC), and mistake injection AOC – while also improving accuracy. Our analysis finds that these gains are robust across training domains and are associated with shorter and more direct CoTs.</abstract>
<identifier type="citekey">sivakumaran-etal-2026-balancing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2140/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>45711</start>
<end>45730</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution
%A Sivakumaran, Nithin
%A Yu, Shoubin
%A Lee, Hyunji
%A Zhang, Yue
%A Payani, Ali
%A Bansal, Mohit
%A Stengel-Eskin, Elias
%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 sivakumaran-etal-2026-balancing
%X Chain-of-thought (CoT) reasoning sometimes fails to faithfully reflect the true computation of a large language model (LLM), hampering its utility in explaining how LLMs arrive at their answers. Moreover, optimizing for faithfulness and interpretability in reasoning often degrades task performance. To address this tradeoff and improve CoT faithfulness, we propose Reasoning Execution by Multiple Listeners (REMuL), a multi-party reinforcement learning approach. REMuL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful. A speaker model generates a reasoning trace, which is truncated and passed to a pool of listener models who “execute” the trace, continuing the trace to an answer. Speakers are rewarded for producing reasoning that is clear to listeners, with additional correctness regularization via masked supervised finetuning to counter the tradeoff between faithfulness and performance. On multiple reasoning benchmarks (BIG-Bench Extra Hard, MuSR, ZebraLogicBench, and FOLIO), REMuL consistently and substantially improves three measures of faithfulness – hint attribution, early answering area over the curve (AOC), and mistake injection AOC – while also improving accuracy. Our analysis finds that these gains are robust across training domains and are associated with shorter and more direct CoTs.
%U https://aclanthology.org/2026.findings-emnlp.2140/
%P 45711-45730
Markdown (Informal)
[Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution](https://aclanthology.org/2026.findings-emnlp.2140/) (Sivakumaran et al., Findings 2026)
ACL