@inproceedings{su-etal-2026-reinforcement,
title = "Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax",
author = "Su, Zeli and
Zhang, Ziyin and
Liu, Zhou and
Song, Xuexian and
Xu, Zhankai and
Zheng, Longfei and
Zhang, Xiaolu and
Fu, Rong and
Xu, Guixian and
Zhang, Wentao",
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.880/",
pages = "17772--17786",
ISBN = "979-8-89176-395-1",
abstract = "Extending large language models (LLMs) to low-resource languages often incurs an ``align- ment tax'': improvements in the target lan- guage come at the cost of catastrophic forget- ting in general capabilities. We argue that this trade-off arises from the rigidity of supervised fine-tuning (SFT), which enforces token-level surface imitation on narrow and biased data distributions. To address this limitation, we propose a semantic-space alignment paradigm powered by Group Relative Policy Optimiza- tion (GRPO), where the model is optimized us- ing embedding-level semantic rewards rather than likelihood maximization. This objective encourages meaning preservation through flex- ible realizations, enabling controlled updates that reduce destructive interference with pre- trained knowledge. We evaluate our approach on Tibetan{--}Chinese machine translation and Ti- betan headline generation. Experiments show that our method acquires low-resource capa- bilities while markedly mitigating alignment tax, preserving general competence more effec- tively than SFT. Despite producing less rigid surface overlap, semantic RL yields higher se- mantic quality and preference in open-ended generation, and few-shot transfer results indi- cate that it learns more transferable and ro- bust representations under limited supervision. Overall, our study demonstrates that reinforce- ment learning with semantic rewards provides a safer and more reliable pathway for inclusive low-resource language expansion."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="su-etal-2026-reinforcement">
<titleInfo>
<title>Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zeli</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziyin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhou</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuexian</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhankai</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Longfei</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaolu</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rong</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guixian</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wentao</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>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>Extending large language models (LLMs) to low-resource languages often incurs an “align- ment tax”: improvements in the target lan- guage come at the cost of catastrophic forget- ting in general capabilities. We argue that this trade-off arises from the rigidity of supervised fine-tuning (SFT), which enforces token-level surface imitation on narrow and biased data distributions. To address this limitation, we propose a semantic-space alignment paradigm powered by Group Relative Policy Optimiza- tion (GRPO), where the model is optimized us- ing embedding-level semantic rewards rather than likelihood maximization. This objective encourages meaning preservation through flex- ible realizations, enabling controlled updates that reduce destructive interference with pre- trained knowledge. We evaluate our approach on Tibetan–Chinese machine translation and Ti- betan headline generation. Experiments show that our method acquires low-resource capa- bilities while markedly mitigating alignment tax, preserving general competence more effec- tively than SFT. Despite producing less rigid surface overlap, semantic RL yields higher se- mantic quality and preference in open-ended generation, and few-shot transfer results indi- cate that it learns more transferable and ro- bust representations under limited supervision. Overall, our study demonstrates that reinforce- ment learning with semantic rewards provides a safer and more reliable pathway for inclusive low-resource language expansion.</abstract>
<identifier type="citekey">su-etal-2026-reinforcement</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.880/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>17772</start>
<end>17786</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax
%A Su, Zeli
%A Zhang, Ziyin
%A Liu, Zhou
%A Song, Xuexian
%A Xu, Zhankai
%A Zheng, Longfei
%A Zhang, Xiaolu
%A Fu, Rong
%A Xu, Guixian
%A Zhang, Wentao
%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 su-etal-2026-reinforcement
%X Extending large language models (LLMs) to low-resource languages often incurs an “align- ment tax”: improvements in the target lan- guage come at the cost of catastrophic forget- ting in general capabilities. We argue that this trade-off arises from the rigidity of supervised fine-tuning (SFT), which enforces token-level surface imitation on narrow and biased data distributions. To address this limitation, we propose a semantic-space alignment paradigm powered by Group Relative Policy Optimiza- tion (GRPO), where the model is optimized us- ing embedding-level semantic rewards rather than likelihood maximization. This objective encourages meaning preservation through flex- ible realizations, enabling controlled updates that reduce destructive interference with pre- trained knowledge. We evaluate our approach on Tibetan–Chinese machine translation and Ti- betan headline generation. Experiments show that our method acquires low-resource capa- bilities while markedly mitigating alignment tax, preserving general competence more effec- tively than SFT. Despite producing less rigid surface overlap, semantic RL yields higher se- mantic quality and preference in open-ended generation, and few-shot transfer results indi- cate that it learns more transferable and ro- bust representations under limited supervision. Overall, our study demonstrates that reinforce- ment learning with semantic rewards provides a safer and more reliable pathway for inclusive low-resource language expansion.
%U https://aclanthology.org/2026.findings-acl.880/
%P 17772-17786
Markdown (Informal)
[Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax](https://aclanthology.org/2026.findings-acl.880/) (Su et al., Findings 2026)
ACL
- Zeli Su, Ziyin Zhang, Zhou Liu, Xuexian Song, Zhankai Xu, Longfei Zheng, Xiaolu Zhang, Rong Fu, Guixian Xu, and Wentao Zhang. 2026. Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax. In Findings of the Association for Computational Linguistics: ACL 2026, pages 17772–17786, San Diego, California, United States. Association for Computational Linguistics.