@inproceedings{zhang-etal-2026-responsibility,
title = "Responsibility-Aware Scheduling for Selective Rationalization",
author = "Zhang, Dengchao and
Zhao, Yunxiao and
Yan, Zhichao and
Fan, Yue and
Guo, Shaoru and
Li, Ru",
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.2584/",
pages = "58731--58746",
ISBN = "979-8-89176-459-0",
abstract = "Select-then-predict rationalizers expose compact rationale-conditioned predictions, but their generator and predictor can interlock: poor rationales mislead the predictor, while predictor lag or shortcut adaptation sends noisy feedback to the generator. Existing studies reduce this interaction by separating generator search from predictor training, but repeated predictor training is expensive, resulting in poor trade-offs. To address this problem, we propose ReSR (Responsibility-Aware Scheduling for Rationalization), a training-control framework that keeps the standard select-then-predict path while adaptively scheduling generator and predictor updates. ReSR uses online, probe, and reference diagnostics with rationale-structure signals to attribute failures to generator-side rationale quality or predictor lag; a virtual-runtime scheduler then converts this attribution into generator/predictor debts and allocates the next optimization slice. Experiments on Toy and HateXplain show that ReSR improves the accuracy{--}fidelity{--}compactness{--}cost balance: ReSR achieves the best Toy highlight F1 and performance comparable to the state-of-the-art approach on HateXplain highlight F1, while attaining stronger classification performance and substantially lower cost."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-responsibility">
<titleInfo>
<title>Responsibility-Aware Scheduling for Selective Rationalization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dengchao</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunxiao</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhichao</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaoru</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ru</namePart>
<namePart type="family">Li</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>Select-then-predict rationalizers expose compact rationale-conditioned predictions, but their generator and predictor can interlock: poor rationales mislead the predictor, while predictor lag or shortcut adaptation sends noisy feedback to the generator. Existing studies reduce this interaction by separating generator search from predictor training, but repeated predictor training is expensive, resulting in poor trade-offs. To address this problem, we propose ReSR (Responsibility-Aware Scheduling for Rationalization), a training-control framework that keeps the standard select-then-predict path while adaptively scheduling generator and predictor updates. ReSR uses online, probe, and reference diagnostics with rationale-structure signals to attribute failures to generator-side rationale quality or predictor lag; a virtual-runtime scheduler then converts this attribution into generator/predictor debts and allocates the next optimization slice. Experiments on Toy and HateXplain show that ReSR improves the accuracy–fidelity–compactness–cost balance: ReSR achieves the best Toy highlight F1 and performance comparable to the state-of-the-art approach on HateXplain highlight F1, while attaining stronger classification performance and substantially lower cost.</abstract>
<identifier type="citekey">zhang-etal-2026-responsibility</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2584/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>58731</start>
<end>58746</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Responsibility-Aware Scheduling for Selective Rationalization
%A Zhang, Dengchao
%A Zhao, Yunxiao
%A Yan, Zhichao
%A Fan, Yue
%A Guo, Shaoru
%A Li, Ru
%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 zhang-etal-2026-responsibility
%X Select-then-predict rationalizers expose compact rationale-conditioned predictions, but their generator and predictor can interlock: poor rationales mislead the predictor, while predictor lag or shortcut adaptation sends noisy feedback to the generator. Existing studies reduce this interaction by separating generator search from predictor training, but repeated predictor training is expensive, resulting in poor trade-offs. To address this problem, we propose ReSR (Responsibility-Aware Scheduling for Rationalization), a training-control framework that keeps the standard select-then-predict path while adaptively scheduling generator and predictor updates. ReSR uses online, probe, and reference diagnostics with rationale-structure signals to attribute failures to generator-side rationale quality or predictor lag; a virtual-runtime scheduler then converts this attribution into generator/predictor debts and allocates the next optimization slice. Experiments on Toy and HateXplain show that ReSR improves the accuracy–fidelity–compactness–cost balance: ReSR achieves the best Toy highlight F1 and performance comparable to the state-of-the-art approach on HateXplain highlight F1, while attaining stronger classification performance and substantially lower cost.
%U https://aclanthology.org/2026.emnlp-main.2584/
%P 58731-58746
Markdown (Informal)
[Responsibility-Aware Scheduling for Selective Rationalization](https://aclanthology.org/2026.emnlp-main.2584/) (Zhang et al., EMNLP 2026)
ACL
- Dengchao Zhang, Yunxiao Zhao, Zhichao Yan, Yue Fan, Shaoru Guo, and Ru Li. 2026. Responsibility-Aware Scheduling for Selective Rationalization. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 58731–58746, Budapest, Hungary. Association for Computational Linguistics.