@inproceedings{huang-etal-2026-pearl,
title = "{PEARL}: Perturbation Efficient Alignment Group Relative Reinforcement Learning",
author = "Huang, Yue and
Zhao, ChenZhuo and
Yang, Zhi and
Zhao, Pu and
Yang, Fangkai and
Wang, Lu and
Liu, Mengqiao and
Lin, Qingwei and
Rajmohan, Saravan and
Zhang, Dongmei",
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.1705/",
pages = "36452--36474",
ISBN = "979-8-89176-462-0",
abstract = "Reinforcement learning with verifiable rewards (RLVR) has become a widely used approach for improving the mathematical reasoning ability of large language models (LLMs). However, recent studies show that LLMs may solve a math problem in its original form but fail when the same problem is reworded without changing its meaning. This suggests that the model has not fully acquired the underlying reasoning ability and instead relies on the surface form of the problem. In this paper, we propose PEARL, an RLVR framework that uses different versions of the same math problem as structured training signals. PEARL first groups each original problem with its checked variants to build semantic perturbation clusters, and then uses partially solved clusters as the active RL training set. PEARL further performs cluster-level RL optimization by comparing model outputs across different versions of the same mathematical problem. Experiments on mathematical reasoning benchmarks show that PEARL achieves strong results against competitive RLVR baselines while using a small active training set. These results suggest that semantically equivalent variants can serve as useful training units for improving mathematical reasoning in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-pearl">
<titleInfo>
<title>PEARL: Perturbation Efficient Alignment Group Relative Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">ChenZhuo</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhi</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pu</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fangkai</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mengqiao</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingwei</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Saravan</namePart>
<namePart type="family">Rajmohan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dongmei</namePart>
<namePart type="family">Zhang</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>Reinforcement learning with verifiable rewards (RLVR) has become a widely used approach for improving the mathematical reasoning ability of large language models (LLMs). However, recent studies show that LLMs may solve a math problem in its original form but fail when the same problem is reworded without changing its meaning. This suggests that the model has not fully acquired the underlying reasoning ability and instead relies on the surface form of the problem. In this paper, we propose PEARL, an RLVR framework that uses different versions of the same math problem as structured training signals. PEARL first groups each original problem with its checked variants to build semantic perturbation clusters, and then uses partially solved clusters as the active RL training set. PEARL further performs cluster-level RL optimization by comparing model outputs across different versions of the same mathematical problem. Experiments on mathematical reasoning benchmarks show that PEARL achieves strong results against competitive RLVR baselines while using a small active training set. These results suggest that semantically equivalent variants can serve as useful training units for improving mathematical reasoning in LLMs.</abstract>
<identifier type="citekey">huang-etal-2026-pearl</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1705/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>36452</start>
<end>36474</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PEARL: Perturbation Efficient Alignment Group Relative Reinforcement Learning
%A Huang, Yue
%A Zhao, ChenZhuo
%A Yang, Zhi
%A Zhao, Pu
%A Yang, Fangkai
%A Wang, Lu
%A Liu, Mengqiao
%A Lin, Qingwei
%A Rajmohan, Saravan
%A Zhang, Dongmei
%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 huang-etal-2026-pearl
%X Reinforcement learning with verifiable rewards (RLVR) has become a widely used approach for improving the mathematical reasoning ability of large language models (LLMs). However, recent studies show that LLMs may solve a math problem in its original form but fail when the same problem is reworded without changing its meaning. This suggests that the model has not fully acquired the underlying reasoning ability and instead relies on the surface form of the problem. In this paper, we propose PEARL, an RLVR framework that uses different versions of the same math problem as structured training signals. PEARL first groups each original problem with its checked variants to build semantic perturbation clusters, and then uses partially solved clusters as the active RL training set. PEARL further performs cluster-level RL optimization by comparing model outputs across different versions of the same mathematical problem. Experiments on mathematical reasoning benchmarks show that PEARL achieves strong results against competitive RLVR baselines while using a small active training set. These results suggest that semantically equivalent variants can serve as useful training units for improving mathematical reasoning in LLMs.
%U https://aclanthology.org/2026.findings-emnlp.1705/
%P 36452-36474
Markdown (Informal)
[PEARL: Perturbation Efficient Alignment Group Relative Reinforcement Learning](https://aclanthology.org/2026.findings-emnlp.1705/) (Huang et al., Findings 2026)
ACL
- Yue Huang, ChenZhuo Zhao, Zhi Yang, Pu Zhao, Fangkai Yang, Lu Wang, Mengqiao Liu, Qingwei Lin, Saravan Rajmohan, and Dongmei Zhang. 2026. PEARL: Perturbation Efficient Alignment Group Relative Reinforcement Learning. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 36452–36474, Budapest, Hungary. Association for Computational Linguistics.