arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2609.28395cs.CLcs.LG

针对翻译任务微调大语言模型:通用遗忘缓解方法无法保持机器翻译特有的指令遵循能力

Fine-Tuning LLMs for Translation: General Forgetting Mitigation Does Not Preserve MT-Specific Instruction Following

Niklas Scholz, David Thulke, Abdallah Nasir, Will Allred, Evgeny Matusov, Hermann Ney

首次发表
浏览论文内容

中文总结 AI 辅助

研究针对翻译微调中的灾难性遗忘,发现通用遗忘缓解方法(如弹性权重巩固)虽能保持通用能力,但无法保持机器翻译特有的指令遵循(如形式、性别、长度控制),仅数据混合方法有效但泛化性有限。

中文摘要 AI 辅助

在平行数据上微调大语言模型可以提高翻译质量,但可能导致灾难性遗忘。缓解方法通常通过在通用基准上的保持度来评估。我们探究这些发现是否适用于机器翻译(MT)微调以及机器翻译特有的指令遵循(MT-IF):即修改翻译的指令,如形式、语法性别和长度控制。我们比较了锚定于辅助数据、模型输出和基础模型参数的方法,首先在Llama 3.2 1B Instruct的筛选研究中进行,然后在Llama 3.1 8B Instruct上使用双向阿拉伯语-英语或西班牙语-英语数据进行微调。弹性权重巩固(Elastic Weight Consolidation)在两个阶段都最能保持通用能力;在8B西班牙语模型上,通用基准的平均得分下降1.7分,而标准微调下降11.0分,但其在形式和语法性别控制上的得分仍接近标准微调。只有使用控制任务示例的数据混合方法能保持这些控制能力,但其增益无法迁移到同一任务的未见提示上。

英文摘要

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) fine-tuning 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.

发表机构

  • AppTek GmbH(AppTek公司)
  • RWTH Aachen University(亚琛工业大学)
  • Applied Science Private University(应用科学私立大学)

机构由 AI 辅助整理,请以论文原文为准。

补充信息

↑