@inproceedings{liang-etal-2026-dual,
title = "{DUAL} {RM}: Beyond Rule-based Preference Reward Modeling via Meta-Reward",
author = "Liang, Xiaobo and
Wang, Wanfu and
Huang, Qipeng and
Ding, Yuyang and
Tang, Zecheng and
Ji, Yixin and
Chen, Qianben and
Zhao, Zhe and
Chen, Kehai and
Li, Juntao and
Zhang, Min",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.1729/",
pages = "37281--37296",
ISBN = "979-8-89176-390-6",
abstract = "The ability to model sparse and underspecified rewards, characteristic of human preferences, is fundamental to scaling Reinforcement Learning (RL). Current preference-based reward modeling largely relies on verifiable rewards, where human-annotated labels define rule-based signals. However, these methods face a fundamental bottleneck we term the Matryoshka Doll Problem: a recursive dependency where each reward verifier requires a meta-verifier, leading to continuous and costly dependence on human annotation. In this work, we propose Dual RM, which couples discriminative and generative reward models (DisRMs and GenRMs) under a non-parametric meta-reward. Rather than verifying the correctness of GenRM{'}s reasoning, the meta-reward evaluates its practical impact on response quality. Specifically, GenRM identifies multi-dimensional evaluation rubrics and iteratively refines the response, while DisRM quantifies the quality shifts induced by each rubric. Furthermore, we implement rubric-based test-time scaling to improve sample efficiency and preference alignment under both DPO and GRPO. Our experiments demonstrate that Dual RM achieves strong performance across major preference benchmarks. Notably, even when trained exclusively on language modality, it exhibits robust cross-modal transfer on Omni-RewardBench."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liang-etal-2026-dual">
<titleInfo>
<title>DUAL RM: Beyond Rule-based Preference Reward Modeling via Meta-Reward</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiaobo</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wanfu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qipeng</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuyang</namePart>
<namePart type="family">Ding</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zecheng</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixin</namePart>
<namePart type="family">Ji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qianben</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhe</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kehai</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Juntao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Min</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>The ability to model sparse and underspecified rewards, characteristic of human preferences, is fundamental to scaling Reinforcement Learning (RL). Current preference-based reward modeling largely relies on verifiable rewards, where human-annotated labels define rule-based signals. However, these methods face a fundamental bottleneck we term the Matryoshka Doll Problem: a recursive dependency where each reward verifier requires a meta-verifier, leading to continuous and costly dependence on human annotation. In this work, we propose Dual RM, which couples discriminative and generative reward models (DisRMs and GenRMs) under a non-parametric meta-reward. Rather than verifying the correctness of GenRM’s reasoning, the meta-reward evaluates its practical impact on response quality. Specifically, GenRM identifies multi-dimensional evaluation rubrics and iteratively refines the response, while DisRM quantifies the quality shifts induced by each rubric. Furthermore, we implement rubric-based test-time scaling to improve sample efficiency and preference alignment under both DPO and GRPO. Our experiments demonstrate that Dual RM achieves strong performance across major preference benchmarks. Notably, even when trained exclusively on language modality, it exhibits robust cross-modal transfer on Omni-RewardBench.</abstract>
<identifier type="citekey">liang-etal-2026-dual</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.1729/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>37281</start>
<end>37296</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DUAL RM: Beyond Rule-based Preference Reward Modeling via Meta-Reward
%A Liang, Xiaobo
%A Wang, Wanfu
%A Huang, Qipeng
%A Ding, Yuyang
%A Tang, Zecheng
%A Ji, Yixin
%A Chen, Qianben
%A Zhao, Zhe
%A Chen, Kehai
%A Li, Juntao
%A Zhang, Min
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F liang-etal-2026-dual
%X The ability to model sparse and underspecified rewards, characteristic of human preferences, is fundamental to scaling Reinforcement Learning (RL). Current preference-based reward modeling largely relies on verifiable rewards, where human-annotated labels define rule-based signals. However, these methods face a fundamental bottleneck we term the Matryoshka Doll Problem: a recursive dependency where each reward verifier requires a meta-verifier, leading to continuous and costly dependence on human annotation. In this work, we propose Dual RM, which couples discriminative and generative reward models (DisRMs and GenRMs) under a non-parametric meta-reward. Rather than verifying the correctness of GenRM’s reasoning, the meta-reward evaluates its practical impact on response quality. Specifically, GenRM identifies multi-dimensional evaluation rubrics and iteratively refines the response, while DisRM quantifies the quality shifts induced by each rubric. Furthermore, we implement rubric-based test-time scaling to improve sample efficiency and preference alignment under both DPO and GRPO. Our experiments demonstrate that Dual RM achieves strong performance across major preference benchmarks. Notably, even when trained exclusively on language modality, it exhibits robust cross-modal transfer on Omni-RewardBench.
%U https://aclanthology.org/2026.acl-long.1729/
%P 37281-37296
Markdown (Informal)
[DUAL RM: Beyond Rule-based Preference Reward Modeling via Meta-Reward](https://aclanthology.org/2026.acl-long.1729/) (Liang et al., ACL 2026)
ACL
- Xiaobo Liang, Wanfu Wang, Qipeng Huang, Yuyang Ding, Zecheng Tang, Yixin Ji, Qianben Chen, Zhe Zhao, Kehai Chen, Juntao Li, and Min Zhang. 2026. DUAL RM: Beyond Rule-based Preference Reward Modeling via Meta-Reward. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 37281–37296, San Diego, California, United States. Association for Computational Linguistics.