@inproceedings{chen-etal-2026-grpo,
title = "{GRPO}-{CARE}: Consistency-Aware Reinforcement Learning for Multimodal Reasoning",
author = "Chen, Yi and
Ge, Yuying and
Wang, Rui and
Ge, Yixiao and
Cheng, Junhao and
Shan, Ying and
Liu, Xihui",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.210/",
pages = "4305--4320",
ISBN = "979-8-89176-395-1",
abstract = "Recent reinforcement learning (RL) approaches, such as outcome-supervised GRPO, have advanced reasoning in Large Language Models (LLMs), yet their adaptation to multimodal LLMs (MLLMs) remains underexplored. Progress has been further limited by the lack of evaluation settings that jointly test perception and reasoning under controlled generalization challenges. To enable such analysis, we present **SEED-Bench-R1**, a structured testbed featuring real-world video tasks and hierarchical evaluation across in-distribution, cross-environment, and cross-environment-task scenarios. Our analysis reveals that standard outcome-supervised GRPO often yields ``logical incoherence''{---}achieving correct answers through flawed reasoning{---}due to its exclusive focus on final-answer rewards and rigid KL penalties. To address this, we propose **GRPO-CARE**, a consistency-aware RL framework that eliminates KL penalties while introducing a two-tiered reward system: a base reward for accuracy and an adaptive bonus for consistency. This bonus, derived from a slowly evolving reference model through group-relative likelihood calibration, rewards reasoning paths that logically support the final answer without requiring expensive process supervision. Experiments on SEED-Bench-R1 show that GRPO-CARE consistently outperforms standard GRPO, achieving a 6.7{\%} gain on the hardest evaluation level and a 24.5{\%} increase in reasoning consistency. Moreover, models trained with GRPO-CARE transfer effectively to diverse video understanding and even language-only reasoning benchmarks, validating its robustness and generality."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-grpo">
<titleInfo>
<title>GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuying</namePart>
<namePart type="family">Ge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixiao</namePart>
<namePart type="family">Ge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junhao</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ying</namePart>
<namePart type="family">Shan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xihui</namePart>
<namePart type="family">Liu</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>Findings of the Association for Computational Linguistics: ACL 2026</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">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Recent reinforcement learning (RL) approaches, such as outcome-supervised GRPO, have advanced reasoning in Large Language Models (LLMs), yet their adaptation to multimodal LLMs (MLLMs) remains underexplored. Progress has been further limited by the lack of evaluation settings that jointly test perception and reasoning under controlled generalization challenges. To enable such analysis, we present **SEED-Bench-R1**, a structured testbed featuring real-world video tasks and hierarchical evaluation across in-distribution, cross-environment, and cross-environment-task scenarios. Our analysis reveals that standard outcome-supervised GRPO often yields “logical incoherence”—achieving correct answers through flawed reasoning—due to its exclusive focus on final-answer rewards and rigid KL penalties. To address this, we propose **GRPO-CARE**, a consistency-aware RL framework that eliminates KL penalties while introducing a two-tiered reward system: a base reward for accuracy and an adaptive bonus for consistency. This bonus, derived from a slowly evolving reference model through group-relative likelihood calibration, rewards reasoning paths that logically support the final answer without requiring expensive process supervision. Experiments on SEED-Bench-R1 show that GRPO-CARE consistently outperforms standard GRPO, achieving a 6.7% gain on the hardest evaluation level and a 24.5% increase in reasoning consistency. Moreover, models trained with GRPO-CARE transfer effectively to diverse video understanding and even language-only reasoning benchmarks, validating its robustness and generality.</abstract>
<identifier type="citekey">chen-etal-2026-grpo</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.210/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>4305</start>
<end>4320</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning
%A Chen, Yi
%A Ge, Yuying
%A Wang, Rui
%A Ge, Yixiao
%A Cheng, Junhao
%A Shan, Ying
%A Liu, Xihui
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F chen-etal-2026-grpo
%X Recent reinforcement learning (RL) approaches, such as outcome-supervised GRPO, have advanced reasoning in Large Language Models (LLMs), yet their adaptation to multimodal LLMs (MLLMs) remains underexplored. Progress has been further limited by the lack of evaluation settings that jointly test perception and reasoning under controlled generalization challenges. To enable such analysis, we present **SEED-Bench-R1**, a structured testbed featuring real-world video tasks and hierarchical evaluation across in-distribution, cross-environment, and cross-environment-task scenarios. Our analysis reveals that standard outcome-supervised GRPO often yields “logical incoherence”—achieving correct answers through flawed reasoning—due to its exclusive focus on final-answer rewards and rigid KL penalties. To address this, we propose **GRPO-CARE**, a consistency-aware RL framework that eliminates KL penalties while introducing a two-tiered reward system: a base reward for accuracy and an adaptive bonus for consistency. This bonus, derived from a slowly evolving reference model through group-relative likelihood calibration, rewards reasoning paths that logically support the final answer without requiring expensive process supervision. Experiments on SEED-Bench-R1 show that GRPO-CARE consistently outperforms standard GRPO, achieving a 6.7% gain on the hardest evaluation level and a 24.5% increase in reasoning consistency. Moreover, models trained with GRPO-CARE transfer effectively to diverse video understanding and even language-only reasoning benchmarks, validating its robustness and generality.
%U https://aclanthology.org/2026.findings-acl.210/
%P 4305-4320
Markdown (Informal)
[GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning](https://aclanthology.org/2026.findings-acl.210/) (Chen et al., Findings 2026)
ACL
- Yi Chen, Yuying Ge, Rui Wang, Yixiao Ge, Junhao Cheng, Ying Shan, and Xihui Liu. 2026. GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning. In Findings of the Association for Computational Linguistics: ACL 2026, pages 4305–4320, San Diego, California, United States. Association for Computational Linguistics.