遗忘
计算机科学
形式
控制(管理)
自然语言处理
人工智能
传输(计算)
要价
质量(理念)
基础(拓扑)
学习迁移
认知心理学
心理学
机器学习
数据质量
语言学
翻译(生物学)
模式(计算机接口)
合并(业务)
改述
作者
Niklas Scholz,David Thulke,Abdallah Nasir,Will Allred,Evgeny Matusov,Hermann Ney
出处
期刊:
日期:2026-01-01
卷期号:: 631-653
标识
DOI:10.18653/2026.wmt-main.36
摘要
Fine-tuning large language models on parallel data improves translation quality but can cause catastrophic forgetting.Mitigation methods are generally evaluated by retention on general benchmarks.We ask whether these findings transfer to machine translation (MT) finetuning and to MT-specific instruction following (MT-IF): instructions that modify a translation, such as formality, grammatical gender, and length control.We compare methods anchored to auxiliary data, to model outputs, and to the base model parameters, first in a screening study with Llama 3.2 1B Instruct, then on Llama 3.1 8B Instruct fine-tuned on bidirectional Arabic-English or Spanish-English data.Elastic Weight Consolidation preserves general capabilities best in both stages; on the 8B Spanish model the average score on general benchmarks drops 1.7 points versus 11.0 for standard fine-tuning, yet its scores for formality and grammatical gender control remain close to standard fine-tuning.Only data mixing with control-task examples preserves these controls, but its gains do not transfer to unseen prompts for the same task.
科研通智能强力驱动
Strongly Powered by AbleSci AI