@inproceedings{su-etal-2026-source,
title = "Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation",
author = "Su, Zeli and
Zhang, Ziyin and
Zewei, Pan and
Liu, Zhou and
Huang, Dingcheng and
Li, Dehan and
Xu, Zhankai and
Zheng, Longfei and
Zhang, Xiaolu and
Zhou, Jun and
Zhang, Wentao",
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.674/",
pages = "15521--15534",
ISBN = "979-8-89176-459-0",
abstract = "Low-resource target-language generation is often limited by scarce parallel data, while high-resource source-language monolingual data is abundant but difficult to use with standard supervised fine-tuning. We propose Source-Grounded Semantic Reinforcement Learning (SG-SRL), a resource-utilization framework that converts source-language monolingual data into cross-lingual semantic supervision. SG-SRL performs reference-free reinforcement learning (RL) on source-language data with a cross-lingual semantic reward, followed by lightweight supervised recovery. Although direct reward optimization induces verbosity-based reward hacking, recovery restores fluency, conciseness, and task format without erasing the learned semantic alignment. On Chinese-to-Thai generation, SG-SRL outperforms cold-start SFT, source-only continued pretraining, and synthetic-data SFT; human preference and training-independent metrics corroborate the gains. Recovery ablations further show that semantic retention and form restoration are jointly controlled by recovery data and intensity. An in-domain Chinese-to-Tibetan experiment shows that the reward interface can use a contrastively trained encoder when a capable multilingual reranker is unavailable, while also delimiting the framework{'}s dependence on reward-model discrimination."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="su-etal-2026-source">
<titleInfo>
<title>Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation</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">Pan</namePart>
<namePart type="family">Zewei</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">Dingcheng</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dehan</namePart>
<namePart type="family">Li</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">Jun</namePart>
<namePart type="family">Zhou</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>2026-10</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>Low-resource target-language generation is often limited by scarce parallel data, while high-resource source-language monolingual data is abundant but difficult to use with standard supervised fine-tuning. We propose Source-Grounded Semantic Reinforcement Learning (SG-SRL), a resource-utilization framework that converts source-language monolingual data into cross-lingual semantic supervision. SG-SRL performs reference-free reinforcement learning (RL) on source-language data with a cross-lingual semantic reward, followed by lightweight supervised recovery. Although direct reward optimization induces verbosity-based reward hacking, recovery restores fluency, conciseness, and task format without erasing the learned semantic alignment. On Chinese-to-Thai generation, SG-SRL outperforms cold-start SFT, source-only continued pretraining, and synthetic-data SFT; human preference and training-independent metrics corroborate the gains. Recovery ablations further show that semantic retention and form restoration are jointly controlled by recovery data and intensity. An in-domain Chinese-to-Tibetan experiment shows that the reward interface can use a contrastively trained encoder when a capable multilingual reranker is unavailable, while also delimiting the framework’s dependence on reward-model discrimination.</abstract>
<identifier type="citekey">su-etal-2026-source</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.674/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>15521</start>
<end>15534</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation
%A Su, Zeli
%A Zhang, Ziyin
%A Zewei, Pan
%A Liu, Zhou
%A Huang, Dingcheng
%A Li, Dehan
%A Xu, Zhankai
%A Zheng, Longfei
%A Zhang, Xiaolu
%A Zhou, Jun
%A Zhang, Wentao
%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 su-etal-2026-source
%X Low-resource target-language generation is often limited by scarce parallel data, while high-resource source-language monolingual data is abundant but difficult to use with standard supervised fine-tuning. We propose Source-Grounded Semantic Reinforcement Learning (SG-SRL), a resource-utilization framework that converts source-language monolingual data into cross-lingual semantic supervision. SG-SRL performs reference-free reinforcement learning (RL) on source-language data with a cross-lingual semantic reward, followed by lightweight supervised recovery. Although direct reward optimization induces verbosity-based reward hacking, recovery restores fluency, conciseness, and task format without erasing the learned semantic alignment. On Chinese-to-Thai generation, SG-SRL outperforms cold-start SFT, source-only continued pretraining, and synthetic-data SFT; human preference and training-independent metrics corroborate the gains. Recovery ablations further show that semantic retention and form restoration are jointly controlled by recovery data and intensity. An in-domain Chinese-to-Tibetan experiment shows that the reward interface can use a contrastively trained encoder when a capable multilingual reranker is unavailable, while also delimiting the framework’s dependence on reward-model discrimination.
%U https://aclanthology.org/2026.emnlp-main.674/
%P 15521-15534
Markdown (Informal)
[Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation](https://aclanthology.org/2026.emnlp-main.674/) (Su et al., EMNLP 2026)
ACL
- Zeli Su, Ziyin Zhang, Pan Zewei, Zhou Liu, Dingcheng Huang, Dehan Li, Zhankai Xu, Longfei Zheng, Xiaolu Zhang, Jun Zhou, and Wentao Zhang. 2026. Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 15521–15534, Budapest, Hungary. Association for Computational Linguistics.