@inproceedings{wu-etal-2025-boosting,
title = "Boosting {LLM} Translation Skills without General Ability Loss via Rationale Distillation",
author = "Wu, Junhong and
Zhao, Yang and
Xu, Yangyifan and
Liu, Bing and
Zong, Chengqing",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.632/",
doi = "10.18653/v1/2025.findings-acl.632",
pages = "12217--12236",
ISBN = "979-8-89176-256-5",
abstract = "Large Language Models (LLMs) have achieved impressive results across numerous NLP tasks, and fine-tuning them for Machine Translation (MT) has improved their performance. However, vanilla fine-tuning often leads to catastrophic forgetting, compromising the broad general abilities of LLMs and introducing potential security risks. These abilities, which are developed using proprietary and unavailable training data, make simple data replay methods ineffective. To overcome this issue, we propose a novel approach called **Ra**tionale **Dis**tillation. RaDis harnesses the strong generative capabilities of LLMs to create rationales for training data, which are then ``replayed'' to prevent forgetting. These rationales connect prior knowledge with new tasks, acting as self-distillation targets to regulate the training process. By jointly training on reference translations and self-generated rationales, the model can learn new translation skills while preserving its general abilities across other tasks. Additionally, RaDis provides a fresh perspective on using rationales in the CL field and has the potential to serve as a general continual learning method for a variety of tasks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wu-etal-2025-boosting">
<titleInfo>
<title>Boosting LLM Translation Skills without General Ability Loss via Rationale Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Junhong</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yangyifan</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bing</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chengqing</namePart>
<namePart type="family">Zong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) have achieved impressive results across numerous NLP tasks, and fine-tuning them for Machine Translation (MT) has improved their performance. However, vanilla fine-tuning often leads to catastrophic forgetting, compromising the broad general abilities of LLMs and introducing potential security risks. These abilities, which are developed using proprietary and unavailable training data, make simple data replay methods ineffective. To overcome this issue, we propose a novel approach called **Ra**tionale **Dis**tillation. RaDis harnesses the strong generative capabilities of LLMs to create rationales for training data, which are then “replayed” to prevent forgetting. These rationales connect prior knowledge with new tasks, acting as self-distillation targets to regulate the training process. By jointly training on reference translations and self-generated rationales, the model can learn new translation skills while preserving its general abilities across other tasks. Additionally, RaDis provides a fresh perspective on using rationales in the CL field and has the potential to serve as a general continual learning method for a variety of tasks.</abstract>
<identifier type="citekey">wu-etal-2025-boosting</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.632</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.632/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>12217</start>
<end>12236</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Boosting LLM Translation Skills without General Ability Loss via Rationale Distillation
%A Wu, Junhong
%A Zhao, Yang
%A Xu, Yangyifan
%A Liu, Bing
%A Zong, Chengqing
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F wu-etal-2025-boosting
%X Large Language Models (LLMs) have achieved impressive results across numerous NLP tasks, and fine-tuning them for Machine Translation (MT) has improved their performance. However, vanilla fine-tuning often leads to catastrophic forgetting, compromising the broad general abilities of LLMs and introducing potential security risks. These abilities, which are developed using proprietary and unavailable training data, make simple data replay methods ineffective. To overcome this issue, we propose a novel approach called **Ra**tionale **Dis**tillation. RaDis harnesses the strong generative capabilities of LLMs to create rationales for training data, which are then “replayed” to prevent forgetting. These rationales connect prior knowledge with new tasks, acting as self-distillation targets to regulate the training process. By jointly training on reference translations and self-generated rationales, the model can learn new translation skills while preserving its general abilities across other tasks. Additionally, RaDis provides a fresh perspective on using rationales in the CL field and has the potential to serve as a general continual learning method for a variety of tasks.
%R 10.18653/v1/2025.findings-acl.632
%U https://aclanthology.org/2025.findings-acl.632/
%U https://doi.org/10.18653/v1/2025.findings-acl.632
%P 12217-12236
Markdown (Informal)
[Boosting LLM Translation Skills without General Ability Loss via Rationale Distillation](https://aclanthology.org/2025.findings-acl.632/) (Wu et al., Findings 2025)
ACL